{ "generated": "2026-09-01", "orgs": [ { "name": "Abundant Intelligences (Indigenous AI)", "slug": "abundant-intelligences-indigenous-ai", "website": "https://www.indigenous-ai.net/abundant/", "city": "Montréal", "province": "QC", "country": "Canada", "sector": "university", "kind": "funder-program", "focus": [ "field-building" ], "status": "", "summary": { "en": "Six-year programme rethinking how AI is conceptualised and designed from Indigenous Knowledge systems: 48 co-investigators and collaborators across 13 universities/research institutes and 8 community organisations in Canada, the US and New Zealand.", "fr": "" }, "key_people": [ "Jason Edward Lewis (co-director", "Concordia Indigenous Futures Research Centre)", "Hemi Whaanga (Massey)" ], "founded": "2023", "founded_year": "2023", "confidence": "secondary", "last_verified": "", "sources": [ "https://www.indigenous-ai.net/abundant/" ], "scale": { "count": 48, "unit": "co-investigators and collaborators", "bucket": "11-50", "safety_count": null, "budget": "", "source": "https://www.indigenous-ai.net/abundant/", "checked": "2026-08-31" }, "notes": "Closes the doc's flagged gap \"Indigenous AI governance voices absent\". This is the main Canadian anchor. Epistemology and design, not catastrophic risk. Traces to CIFAR's 2019 Indigenous-perspectives position paper.", "editorial_scope": "adjacent", "dedicated": false }, { "name": "Aether AI Safety Research", "slug": "aether-ai-safety-research", "website": "https://aether-ai-research.org", "city": "Toronto", "province": "ON", "country": "Canada", "sector": "", "kind": "research-lab", "focus": [ "alignment-control", "evaluations", "interpretability" ], "status": "", "summary": { "en": "Independent technical AI safety group. Own words: \"Aether is an AI safety research group. Our primary focus is on studying LLM monitoring and hidden reasoning, but we are also interested in understanding and shaping the propensities of LLMs.\" Work to date is chain-of-thought monitoring (how information access affects a monitor's ability to detect sabotage, ICML 2026), a taxonomy of hidden reasoning in LLMs, and a no-CoT task-completion time-horizon benchmark. Current focus is a \"science of held-out monitors\": when held-out measures of alignment silently degrade under training against correlated proxies, and when they remain trustworthy.", "fr": "" }, "key_people": [ "Rauno Arike (Managing Director & Research Scientist)", "Shubhorup Biswas", "Dennis Akar", "Vladimir Ivanov", "Joey Yudelson (Research Scientists)", "advisors: Rohan Subramani (founder; now at Coefficient Giving)", "Seth Herd (Astera Institute)", "Zhijing Jin (University of Toronto)", "Francis Rhys Ward (Arrow Research)" ], "founded": "", "founded_year": "", "confidence": "AI verified", "last_verified": "2026-08-31", "sources": [ "https://aether-ai-research.org/assets/index-BZfejIrb.js" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "https://aether-ai-research.org/assets/index-BZfejIrb.js", "checked": "2026-08-31" }, "notes": "Site renders client-side from a GitHub Pages shell (empty div#root), so all content was read from its React bundle, 2026-08-31; re-fetch the bundle to check any quote here. CANADIAN CONNECTION ESTABLISHED FIRST-PARTY, not inherited from AIGS: Aether's own hiring page states \"Location: Trajectory Labs, Toronto (in-person expected, visa sponsorship available)\", and Trajectory Labs is already a row here. Corroborated by founder Rohan Subramani's own site (\"Previously, I founded Aether... I am also on leave from my PhD at the University of Toronto, where I am supervised by Professor Zhijing Jin\"), plus an October 2025 Trajectory Labs talk with Rauno Arike. ⚠️ FIELD PROVENANCE: City/Province come from Aether's own CLOSED hiring round (applications shut 2026-01-17, page marked \"preserved for reference\"), NOT from the homepage, which states no location. TWO CORROBORATION CHECKS CAME BACK EMPTY on 2026-08-31 and are recorded because a positive would have been quoted: grep for \"aether\" over trajectorylabs.org /, /about-us, /programs and /events returned 0 hits on all four, so the host org does not name them back; and Arike's LessWrong profile states no location or affiliation. There is therefore NO first-party evidence of where Aether operates TODAY as distinct from where the closed hiring round said it would. IF THAT WEAKENS, BLANK City AND KEEP THE CLAIM HERE. Sector is left EMPTY rather than inferred: no incorporation, charity number or legal jurisdiction appears anywhere on the site. FUNDING: \"We are funded by Coefficient Giving\"; amount not published. No headcount figure is published; five researcher and four advisor cards are not a scale figure under this map's rule. CORRECTION to prior-art.md §1.2: the site does NOT claim collaborations with Redwood, MATS and Astra — MATS appears only as a hiring benchmark and Redwood only as an aspirational comparison.", "editorial_scope": "core", "dedicated": true }, { "name": "AI Alignment McGill", "slug": "ai-alignment-mcgill", "website": "https://www.alignmentforum.org/groups/TLbuSuKccZThdvDBo", "city": "Montréal", "province": "QC", "country": "Canada", "sector": "student group", "kind": "student-group", "focus": [ "alignment-control", "field-building" ], "status": "", "summary": { "en": "Reading groups, hackathons and related alignment/safety activities.", "fr": "" }, "key_people": [ "Organisers listed only by Alignment Forum handles (BionicD0LPH1N", "Fraser", "TheBayesian", "jingxiangmo)" ], "founded": "", "founded_year": "", "confidence": "secondary", "last_verified": "", "sources": [ "https://www.alignmentforum.org/groups/TLbuSuKccZThdvDBo" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "PARTIALLY RESOLVED 2026-07-29: contact and Discord (discord.gg/C3p5eMkuwV) found; real-name leadership still unknown and no independent site exists. ACTIVITY SIGNALS CONFLICT: its Alignment Forum group page lists no upcoming and no past events, which reads dormant, while aisafety.com/communities lists it as Active. Treat as uncertain until someone checks the Discord.", "editorial_scope": "pipeline", "dedicated": true }, { "name": "AI Atlantic Research Consortium", "slug": "ai-atlantic-research-consortium", "website": "https://aiarc.ca", "city": "Halifax (Atlantic region)", "province": "NS", "country": "Canada", "sector": "consortium", "kind": "network", "focus": [ "ethics-societal", "education" ], "status": "", "summary": { "en": "Collaborative AI leadership group across Atlantic Canada connecting research, academia, industry and public outreach on responsible AI.", "fr": "" }, "key_people": [], "founded": "2024-10", "founded_year": "2024", "confidence": "secondary", "last_verified": "", "sources": [ "https://aiarc.ca/" ], "scale": { "count": 9, "unit": "member institutions (NOT a headcount)", "bucket": "3-10", "safety_count": null, "budget": "", "source": "https://aiarc.ca/", "checked": "2026-08-31" }, "notes": "Best available Atlantic coverage. NOT safety-specific, so the doc's \"Atlantic coverage thin\" finding stands.", "editorial_scope": "adjacent", "dedicated": false }, { "name": "AI Safety Canada Meetup (BlueDot Impact)", "slug": "ai-safety-canada-meetup-bluedot-impact", "website": "https://luma.com/bluedotevents", "city": "Online (national)", "province": "Country-wide", "country": "Canada", "sector": "nonprofit", "kind": "community", "focus": [ "field-building" ], "status": "", "summary": { "en": "Free monthly online meetup (Zoom, ~2h) built as the accessible entry point to AI safety for a general Canadian audience, and explicitly designed to hand people onward to local in-person orgs. Lightning talks plus breakout discussions; a shared participant-intro doc circulates at each event.", "fr": "" }, "key_people": [ "Cinthia Nemoto (organiser)", "volunteer moderators and speakers" ], "founded": "2025-10 (first meetup); 2026-01 (BlueDot partnership)", "founded_year": "2025", "confidence": "secondary", "last_verified": "", "sources": [], "scale": { "count": 43, "unit": "registrants for one event (2026-09-02), not members", "bucket": "11-50", "safety_count": null, "budget": "", "source": "https://luma.com/bluedotevents", "checked": "2026-08-31" }, "notes": "Resolves this sheet's open \"AIGS Slack\" question: this meetup, not the AIGS Slack, is the most likely referent of the \"Canadian AI safety monthly calls\". Positions itself as a lead generator for AIGS, Trajectory Labs and HΩ. Paused for July and August 2026; next event 2 Sep 2026. Grant application submitted 2026-04 with an HΩ support letter. BASIS: primary HΩ correspondence and Luma event mail, not a fetched website, hence secondary under this sheet's definition.", "editorial_scope": "pipeline", "dedicated": false }, { "name": "AIGS Canada (AI Governance & Safety Canada)", "slug": "aigs-canada-ai-governance-safety-canada", "website": "https://aigs.ca", "city": "Ottawa", "province": "ON", "country": "Canada", "sector": "nonprofit", "kind": "community", "focus": [ "governance-policy", "field-building", "education" ], "status": "", "summary": { "en": "Canadian AI-safety advocacy: white papers, legislative recommendations, government submissions, 7+ parliamentary appearances, fellowships, policy courses, and a ~700-person Slack community.", "fr": "" }, "key_people": [ "Wyatt Tessari L'Allié (founder / ED)" ], "founded": "2022", "founded_year": "2022", "confidence": "AI verified", "last_verified": "", "sources": [ "https://aigs.ca/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Its 2025 white paper explicitly warns of smarter-than-human AI within 1-3 years and a \"looming global crisis\". Office at 123 Slater St #610, Ottawa. OPEN QUESTION: whether its Slack is the \"Canadian AI safety monthly calls\" community. FIELD PROVENANCE: aigs.ca states neither a founding year nor its lead: the 2022 date and Wyatt Tessari L'Allié as founder/ED are inherited from the HΩ landscape doc.", "editorial_scope": "core", "dedicated": false }, { "name": "Amii: AI trust & safety", "slug": "amii-ai-trust-safety", "website": "https://amii.ca", "city": "Edmonton", "province": "AB", "country": "Canada", "sector": "research institute", "kind": "university-institute", "focus": [ "security-robustness" ], "status": "", "summary": { "en": "Programme stood up Jan 2025; reinforcement-learning depth applied to robustness and safety. CAISI partner.", "fr": "" }, "key_people": [ "Alyssa Lefaivre Škopac (Director)", "Linglong Kong", "Bei Jiang" ], "founded": "2025-01 (programme)", "founded_year": "2025", "confidence": "unverified (doc)", "last_verified": "", "sources": [ "https://www.amii.ca/about/canadian-ai-safety-institute" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "One of Canada's three national AI institutes. Two Amii Fellows won CIFAR / UK-AISI Alignment Project grants.", "editorial_scope": "core", "dedicated": false }, { "name": "Armilla AI", "slug": "armilla-ai", "website": "https://www.armilla.ai", "city": "Toronto", "province": "ON", "country": "Canada", "sector": "company", "kind": "company", "focus": [ "evaluations", "security-robustness", "governance-policy" ], "status": "", "summary": { "en": "Lloyd's-backed managing general agent underwriting AI liability: affirmative AI liability coverage, AI performance warranties, and AI model risk assessments. Its assessment arm runs \"AI/LLM testing, red-teaming, and third-party evaluations\", scored on accuracy, fairness and resilience against the EU AI Act, Colorado AI Act, NIST AI RMF and ISO 42001, issuing an \"Armilla Verified Seal\".", "fr": "" }, "key_people": [ "Karthik Ramakrishnan (CEO)", "Philip Dawson (Head of Distribution & Partnerships)", "Geoff MacGregor (Head of Strategy and Operations)", "Andrew Correll (Director of Underwriting)", "Alec Eyre (Head of AI Assessments)" ], "founded": "", "founded_year": "", "confidence": "AI verified", "last_verified": "2026-08-31", "sources": [ "https://www.armilla.ai/company" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "https://www.armilla.ai/company", "checked": "2026-08-31" }, "notes": "Own site fetched 2026-08-31. SCOPE CALL — adjacent, not core, and this is deliberate. 80,000 Hours tags Armilla roles \"AI safety\"; that is the job board's editorial judgment applied to an assurance and insurance company, not Armilla's framing. Armilla does say \"Accelerating Humanity's Transition to AI Safely\", but the object of the safety is commercial liability, model performance and bias. The words alignment, existential, catastrophic, frontier and loss-of-control appear NOWHERE on the site. Same bucket as TrojAI and SecuritAI, already adjacent here. LOCATION is first-party but buried in the legal terms, not on any customer-facing page — armilla.ai/terms-of-service, verbatim: \"We are registered in Ontario, Canada and have our registered office at 1179 King Street West, Suite 112, Toronto, On, Canada, M6K 3C5.\" WHY CARRY IT ANYWAY: third-party evaluation and red-teaming capacity physically in Toronto, and the only AI-liability-transfer mechanism anywhere in this map. \"Backed by A-rated global insurers including Chaucer Group, Axis Capital, Convex, Swiss Re, and Greenlight Re\"; footer: \"Armilla Insurance Services is a Coverholder at Lloyd's.\" Those are capacity providers, not funding, which is why the budget field is empty. No headcount is stated; five leadership cards is not a scale figure. LOOSE END: jobs.80000hours.org/organisations/armilla-ai/ returned HTTP 404 on 2026-08-31, so the job board's own tags were not read first-party; the verdict above does not rest on them.", "editorial_scope": "adjacent", "dedicated": true }, { "name": "CAISI (Canadian AI Safety Institute)", "slug": "caisi-canadian-ai-safety-institute", "website": "https://ised-isde.canada.ca/site/ised/en/canadian-artificial-intelligence-safety-institute", "city": "Ottawa", "province": "ON", "country": "Canada", "sector": "government", "kind": "government", "focus": [ "evaluations", "field-building" ], "status": "", "summary": { "en": "Canada's national AI safety institute: tracking emerging AI risks, transparent model evaluations, and funding safety research. Member of the International Network of AI Safety Institutes. Functions largely as a funder and coordinator rather than a research producer.", "fr": "" }, "key_people": [ "No individual head named publicly: CAISI is led by ISED as a department", "and the ISED CAISI page names no Executive Director", "chair or other individual head. The CAISI Research Program at CIFAR (a separate row in this sheet) is co-directed by Catherine Régis (UdeM / Mila) and Nicolas Papernot (U of T / Vector)" ], "founded": "2024-11", "founded_year": "2024", "confidence": "AI verified", "last_verified": "", "sources": [ "https://ised-isde.canada.ca/site/ised/en/canadian-artificial-intelligence-safety-institute" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Announced 12 Nov 2024; $50M over 5 years from Budget 2024's $2.4B AI package. The \"AI for All\" national strategy (4 Jun 2026) commits a FURTHER $50M to expand CAISI's risk tracking and model evaluations. Also reachable at aisafety.ca. FIELD PROVENANCE (fetched 2026-08-28): the ISED CAISI page names no ED, chair or individual head; it says CAISI is \"led by Innovation, Science and Economic Development Canada\", drawing on NRC research capability and on the broader Canadian research community through CIFAR. The Régis / Papernot co-director names come from cifar.ca/ai/caisi (fetched 2026-08-28), not from ISED. The budget figures above ($50M over 5 years, and the further $50M in AI for All) are inherited from the 2026-07-29 pass and were NOT re-fetched this pass. CORRECTION 2026-08-28: Key people previously read \"Elissa Strome (Executive Director, appointed Feb 2025)\". That was wrong. Strome is Executive Director of the Pan-Canadian AI Strategy at CIFAR (cifar.ca/bios/elissa-strome, fetched 2026-08-28); neither the ISED CAISI page nor her CIFAR bio connects her to CAISI. The same error is flagged and restated in the HΩ landscape doc 1l-nbGhe0LmWEWC76S4gDr0Hi-1N9trmrG1Ywqvu9slU (Part 2 governance table).", "editorial_scope": "core", "dedicated": true }, { "name": "Carleton SCS: CRC in AI (multi-agent RL safety)", "slug": "carleton-scs-crc-in-ai-multi-agent-rl-safety", "website": "https://carleton.ca/scs/people/sriram-subramanian/", "city": "Ottawa", "province": "ON", "country": "Canada", "sector": "university", "kind": "research-chair", "focus": [ "unclassified" ], "status": "", "summary": { "en": "Multi-agent reinforcement learning in which safety is one of the named problems: his own site frames the work as \"scale, non-stationarity, effective communication, safety, and sample inefficiency in multi-agent learning systems\", and Carleton's own SCS profile lists \"AI Safety\" among his seven research interests. One postdoc opening is advertised at the intersection of multi-agent systems, AI safety and reinforcement learning.", "fr": "" }, "key_people": [ "Sriram Ganapathi Subramanian (Assistant Professor", "School of Computer Science", "Canada Research Chair Tier II in Artificial Intelligence)" ], "founded": "", "founded_year": "", "confidence": "AI verified", "last_verified": "", "sources": [ "https://sriramsubramanian.com/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Started at Carleton July 2025, after a Vector Institute distinguished postdoc (Sep 2022 - Jun 2025); PhD Waterloo 2022, Best Doctoral Dissertation, Canadian AI Association 2023. SCOPE CALL: recorded core rather than adjacent, on the same judgment already made for the SFU Trustworthy AI Lab row - technical safety work that is not framed in catastrophic-risk or alignment terms. His publication list (29 items on his own site, checked 2026-08-28) is MARL-general; the safety-adjacent strands are inverse constrained RL (ICML 2024, TMLR 2024 survey) and mixed-motive LLM-agent evaluation (Concordia, NeurIPS 2025 D&B). No alignment or catastrophic-risk paper. Flip this cell to adjacent if that stays true. NOT INDEPENDENT CAPACITY: he is a faculty affiliate of both the Vector Institute (row 4) and the Schwartz Reisman Institute (row 6). There is no dedicated AI safety lab or centre at Carleton - see the Carleton AI safety presence row in the excluded block. FIELD PROVENANCE: title, chair and tier, research interests and affiliations from Carleton's own SCS profile; the postdoc opening, the address (Herzberg Laboratories HP5360, 1125 Colonel By Drive) and the publication list from his personal site sriramsubramanian.com, both fetched 2026-08-28.", "editorial_scope": "core", "dedicated": false }, { "name": "Carleton SPPA: CRC in Governance and AI", "slug": "carleton-sppa-crc-in-governance-and-ai", "website": "https://carleton.ca/sppa/people/ojo-adegboyega/", "city": "Ottawa", "province": "ON", "country": "Canada", "sector": "university", "kind": "research-chair", "focus": [ "alignment-control", "governance-policy", "ethics-societal" ], "status": "", "summary": { "en": "Governance mechanisms for trustworthy AI-enabled public services: stakeholder expectations as governance requirements, transparency, accountability and organisational trustworthiness in public-sector AI adoption. Public-administration governance of deployed AI, not catastrophic risk or alignment.", "fr": "" }, "key_people": [ "Adegboyega Ojo (Professor", "School of Public Policy and Administration", "Canada Research Chair Tier 1 in Governance and Artificial Intelligence)" ], "founded": "", "founded_year": "", "confidence": "AI verified", "last_verified": "", "sources": [ "https://carleton.ca/sppa/people/ojo-adegboyega/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Clean fit to the adjacent definition (responsible AI and general AI policy; real work, different objective). Chair announced 2022. FIELD PROVENANCE: name, title, Tier 1 and the chair title from Carleton's own SPPA profile, fetched 2026-08-28; the 2022 announcement date and the research-programme summary from carleton.ca newsroom and the federal chairs-chaires directory listing, search-level only this pass. The brief that prompted this row had the tier unstated and low confidence; Carleton's own page resolves it to Tier 1.", "editorial_scope": "adjacent", "dedicated": true }, { "name": "CEIMIA", "slug": "ceimia", "website": "https://ceimia.org", "city": "Montréal", "province": "QC", "country": "Canada", "sector": "nonprofit", "kind": "university-institute", "focus": [ "unclassified" ], "status": "", "summary": { "en": "Applied responsible-AI centre running projects across four programmes: AI for the Environment, AI for Global Health, Data for AI, and Governance and Human Rights. Self-description: \"a key player in the responsible development of artificial intelligence\" focused on \"ethics, human rights, inclusion, diversity, innovation and economic growth\", built around the OECD Principles on Artificial Intelligence. Collaborates within GPAI's working groups and hosts the Montréal hub of the Network of Centres for International Cooperation in AI.", "fr": "" }, "key_people": [ "Sophie Fallaha (Executive Director)", "Stephanie King (Director of AI Initiatives)", "Noémie Gervais (Manager, Montreal Hub, GPAI Network of Centres)", "Philippe Beaudoin (Senior Research Director)" ], "founded": "", "founded_year": "", "confidence": "AI verified", "last_verified": "2026-08-31", "sources": [ "https://ceimia.org/en/about-us/" ], "scale": { "count": 15, "unit": "member organizations", "bucket": "11-50", "safety_count": 0, "budget": "$7M", "source": "https://ceimia.org/en/about-us/", "checked": "2026-08-31" }, "notes": "Corrected 2026-08-31: the \"GPAI centre of expertise\" framing is stale as an identity; CEIMIA's own site no longer leads with it and GPAI now appears as one relationship among several. Safety count is genuinely ZERO rather than \"not distinguished\": all 17 ongoing applied projects sit in the four adjacent programmes, none is alignment, evaluations, control or catastrophic risk.", "editorial_scope": "adjacent", "dedicated": false }, { "name": "CIFAR AI Safety Solution Networks", "slug": "cifar-ai-safety-solution-networks", "website": "https://cifar.ca/ai/cifar-solution-networks/", "city": "Toronto (national)", "province": "ON", "country": "Canada", "sector": "nonprofit", "kind": "funder-program", "focus": [ "ethics-societal", "field-building" ], "status": "", "summary": { "en": "Two networks launched Nov 2025: (1) synthetic / AI-generated evidence infiltrating the Canadian justice system; (2) dialect bias and safer models for linguistic minorities.", "fr": "" }, "key_people": [ "Ebrahim Bagheri (U of T)", "Maura Grossman (Waterloo / Osgoode / Vector)", "Karen Eltis (uOttawa)", "Vered Shwartz (UBC)" ], "founded": "2025-11", "founded_year": "2025", "confidence": "secondary", "last_verified": "", "sources": [ "https://cifar.ca/cifarnews/2025/11/19/cifar-launches-new-ai-safety-solution-networks/" ], "scale": { "count": 12, "unit": "affiliated researchers", "bucket": "11-50", "safety_count": 12, "budget": "$700,000 per network over two years ($1.4M across the two AI Safety networks), 2025-11-19", "source": "https://cifar.ca/cifarnews/2025/11/19/cifar-launches-new-ai-safety-solution-networks/", "checked": "2026-08-31" }, "notes": "This is the main confirmed uOttawa link into Canadian AI safety work. The doc listed uOttawa as an unresolved gap.", "editorial_scope": "core", "dedicated": false }, { "name": "CIFAR: CAISI Research Program", "slug": "cifar-caisi-research-program", "website": "https://cifar.ca/ai/caisi/", "city": "Toronto", "province": "ON", "country": "Canada", "sector": "nonprofit", "kind": "funder-program", "focus": [ "alignment-control", "interpretability", "security-robustness" ], "status": "", "summary": { "en": "Delivers roughly $27M of CAISI's research budget: investigator-led Catalyst Grants (first round 2025, projects starting 2026) on interpretability, robustness testing and cyber-misuse, plus $1M into the UK-AISI-led global Alignment Project.", "fr": "" }, "key_people": [ "Nicolas Papernot & Catherine Régis (research co-directors)" ], "founded": "2025", "founded_year": "2025", "confidence": "AI verified", "last_verified": "", "sources": [ "https://cifar.ca/ai/caisi/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Draws on all three national AI institutes (Amii, Mila, Vector). 2025 year-in-review at report.buildingsafeai.ca. FIELD PROVENANCE (fetched 2026-08-28): the Papernot / Régis co-director titles are now CONFIRMED on cifar.ca/ai/caisi, which lists both as Co-Director of the CAISI Research Program (Papernot: Canada CIFAR AI Chair, Vector; Régis: Canada CIFAR AI Chair, Mila). This supersedes the 2026-07-29 note that the names were inherited unconfirmed from the HΩ landscape doc. The $27M share of the CAISI budget is still inherited from that pass and was NOT re-fetched.", "editorial_scope": "core", "dedicated": true }, { "name": "CIGI: Global AI Risks Initiative", "slug": "cigi-global-ai-risks-initiative", "website": "https://www.cigionline.org/programs/global-ai-risks-initiative/", "city": "Waterloo", "province": "ON", "country": "Canada", "sector": "nonprofit", "kind": "funder-program", "focus": [ "alignment-control", "security-robustness", "governance-policy" ], "status": "", "summary": { "en": "International governance of global and catastrophic AI risk: the proposed Framework Convention on Global AI Challenges, AI national-security scenarios, strategic foresight.", "fr": "" }, "key_people": [ "Duncan Cass-Beggs (Executive Director)" ], "founded": "", "founded_year": "", "confidence": "secondary", "last_verified": "", "sources": [ "https://www.cigionline.org/programs/global-ai-risks-initiative/" ], "scale": { "count": 2, "unit": "FTE staff", "bucket": "1-2", "safety_count": 4, "budget": "", "source": "https://www.cigionline.org/programs/global-ai-risks-initiative/", "checked": "2026-08-31" }, "notes": "MISSING from the HΩ landscape doc's org list (Cass-Beggs appeared only as a person). cigionline.org returns 403 to automated fetch, so details come from CIGI pages via search.", "editorial_scope": "core", "dedicated": true }, { "name": "Cohere / Cohere Labs", "slug": "cohere-cohere-labs", "website": "https://cohere.com", "city": "Toronto", "province": "ON", "country": "Canada", "sector": "company", "kind": "company", "focus": [ "evaluations", "security-robustness", "ethics-societal" ], "status": "", "summary": { "en": "Cohere Labs, the company's ML research lab (established June 2022 as \"Cohere For AI\", renamed \"Cohere Labs\" in April 2025), works toward \"trustworthy, collaborative, and universally capable AI systems\", explicitly \"Advancing AI safety to ensure ML innovation aligns with societal values\", plus open-source safety and multilingual/underserved-language research (Aya, 70+ languages). On the product side Cohere runs adversarial testing of its own systems: \"We simulate adversarial attacks on our systems to test for vulnerabilities and improve defenses\", alongside a bug bounty.", "fr": "" }, "key_people": [ "Aidan Gomez (Cohere)" ], "founded": "2019 / 2022", "founded_year": "2019", "confidence": "AI verified", "last_verified": "2026-08-31", "sources": [ "https://cohere.com/research", "https://cohere.com/security" ], "scale": { "count": 4500, "unit": "community members (Cohere Labs open-science community, self-selecting)", "bucket": "200+", "safety_count": null, "budget": "", "source": "https://cohere.com/research", "checked": "2026-08-31" }, "notes": "NEGATIVE FINDING, 2026-08-31: the \"Responsibility Council\" this entry previously named COULD NOT BE FOUND on cohere.com. Neither cohere.com/research nor cohere.com/security mentions such a body; the phrase came from a third-party directory. Do not restore it without a Cohere page. Cohere does NOT state nonprofit status for Cohere Labs anywhere on its research page, so the previous \"nonprofit research arm\" was unsourced. Product-safety scope, not catastrophic risk. SCALE UNIT: the 4,500 figure is the Cohere Labs Open Science Community, self-selecting participants from 150 countries — structurally the same kind of number as AIGS Canada's 700-person Slack. It must never sit beside an employee count; Cohere's own headcount is not published.", "editorial_scope": "adjacent", "dedicated": true }, { "name": "Concordia Applied AI Institute", "slug": "concordia-applied-ai-institute", "website": "https://www.concordia.ca/research/applied-ai-institute.html", "city": "Montréal", "province": "QC", "country": "Canada", "sector": "university", "kind": "university-institute", "focus": [ "ethics-societal" ], "status": "", "summary": { "en": "AI research clusters including trustworthy/responsible AI.", "fr": "" }, "key_people": [], "founded": "", "founded_year": "", "confidence": "secondary", "last_verified": "", "sources": [ "https://www.concordia.ca/research/applied-ai-institute/research/clusters-centres.html" ], "scale": { "count": 105, "unit": "faculty (plus 200+ graduate students)", "bucket": "51-200", "safety_count": null, "budget": "", "source": "https://www.concordia.ca/research/applied-ai-institute.html", "checked": "2026-08-31" }, "notes": "CHECKED: no dedicated catastrophic-risk safety group at Concordia.", "editorial_scope": "adjacent", "dedicated": false }, { "name": "ControlAI (Canada campaign)", "slug": "controlai-canada-campaign", "website": "https://controlai.org", "city": "", "province": "Country-wide", "country": "Canada", "sector": "nonprofit", "kind": "other", "focus": [ "governance-policy" ], "status": "", "summary": { "en": "Self-description: \"a non-profit organisation that works to prevent the extinction risk posed by superintelligence and to secure a great future for humanity.\" Its Canada campaign asks that Canada \"negotiate an international 'trust but verify' regime to prohibit the development of superintelligent AI\", framed as a national security priority. Delivery in Canada is parliamentary briefings, committee testimony (RIDR, TRCM, ETHI, INDU), and a published Canada Campaign Statement carrying signatures from parliamentarians across every party.", "fr": "" }, "key_people": [ "Samuel Buteau (leads the Canadian work; previously Senior AI Safety Scientist at Mila)", "Andrea Miotti (Founder & CEO)", "Connor Leahy (US Executive Director)" ], "founded": "", "founded_year": "", "confidence": "AI verified", "last_verified": "2026-08-31", "sources": [ "https://controlai.org/our-work-in-canada", "https://controlai.org/canada-statement/en" ], "scale": { "count": 2327, "unit": "community members", "bucket": "200+", "safety_count": null, "budget": "", "source": "https://discord.com/api/v9/invites/6trcsqEaYS?with_counts=true", "checked": "2026-08-31" }, "notes": "Own site fetched 2026-08-31. Surfaced while pulling scale figures for the MASEG meetup, which listed a past event \"Introducing ControlAI Canada\" (2026-08-13). controlai.com 301-redirects to controlai.org; use the .org. ⚠️ CLASSIFICATION CALL — THERE IS NO CANADIAN LEGAL ENTITY. controlai.org/about, verbatim: \"In the UK, we are a not-for-profit company limited by guarantee. In the US, we are a 501(c)(4) social welfare organization.\" Canada is named as a country of activity, not of incorporation. This is the MIRROR IMAGE of the Evitable row: Evitable is Canadian-led but US-scoped on a US fiscal sponsor; ControlAI Canada is foreign-incorporated but Canada-scoped with a Canada-based lead. Both belong in the map with the asymmetry stated. The country value follows gaps-and-discovery.md's wording; the map has no settled convention for this shape yet. ACHIEVEMENTS, first-party but SELF-REPORTED and labelled as such: \"briefed more than 120 Canadian MPs and Senators on the extinction risk posed by superintelligent AI\" in roughly nine months; 30+ lawmakers publicly supporting; testimony before Senate and House committees (RIDR, TRCM, ETHI, INDU); in February 2026 Buteau alone briefed 34 Canadian lawmakers. CANADA CAMPAIGN STATEMENT, published 2026-08-28 — the same day as the previous research pass, which is why it was missed — carries 15+ named Canadian parliamentarians across every party, including Hon. Judy A. Sgro MP, Hon. Jonathan Wilkinson MP, Hon. Steven Guilbeault MP, Joël Godin MP, Cathay Wagantall MP, Martin Champoux MP (for the Bloc caucus) and Senators Colin Deacon, Kim Pate, Tony Loffreda and Julie Miville-Dechêne. ⚠️ CITY NOT RECORDED: no city or province is stated on their site. Buteau's Mila background implies Montréal but that is an inference, not a source. NOT IN THE aisafety.com CANADIAN SET: ControlAI appears there only as a global online Discord with location null, a genuine gap in aisafety.com and not just in this map. SCALE: 2,327 Discord members with 205 online, ORG-WIDE and not Canada, from Discord's public unauthenticated invite endpoint (no server joined). Staff headcount is not published as a figure — the about page names roughly 18 team members and 2 advisors but states no total. Canada specifically has 1 named lead. No funding figure on the about page or the Canada page. LEGAL FORM: ControlAI is UK/US-incorporated with no Canadian legal entity; this entry covers its Canada-scoped campaign. Recorded here rather than in Country, which is a structured field and rendered \"Country-wide, UK/US-incorporated, Canada-scoped campaign\" on the page. The map still has no convention for orgs whose legal and operating footprints differ (ControlAI, Evitable, TrojAI, Aether).", "editorial_scope": "core", "dedicated": false }, { "name": "Dalhousie A2I2", "slug": "dalhousie-a2i2", "website": "https://projects.cs.dal.ca/a2i2/", "city": "Halifax", "province": "NS", "country": "Canada", "sector": "university", "kind": "research-lab", "focus": [ "alignment-control" ], "status": "", "summary": { "en": "AI/agents lab; human-AI teaming under pressure via command-and-control simulations (with Laval, DRDC, Thales, NSERC Alliance).", "fr": "" }, "key_people": [], "founded": "", "founded_year": "", "confidence": "secondary", "last_verified": "", "sources": [ "https://projects.cs.dal.ca/a2i2/" ], "scale": { "count": 50, "unit": "affiliated faculty (plus 100+ students and postdocs), as of the 2026-08-20 relaunch", "bucket": "11-50", "safety_count": null, "budget": "", "source": "https://www.dal.ca/news/2026/08/20/atlantic-ai-institute-dalhousie.html", "checked": "2026-08-31" }, "notes": "Defence and human-factors framing, not catastrophic risk.", "editorial_scope": "adjacent", "dedicated": true }, { "name": "Evitable", "slug": "evitable", "website": "https://evitable.com", "city": "Montréal", "province": "QC", "country": "Canada", "sector": "nonprofit", "kind": "other", "focus": [ "unclassified" ], "status": "", "summary": { "en": "Public-facing campaigning to inform and organise the public against the reckless race to superintelligence, moving the powerful-AI conversation out of technical circles.", "fr": "" }, "key_people": [ "David Krueger (founder)" ], "founded": "2025", "founded_year": "2025", "confidence": "secondary", "last_verified": "", "sources": [ "https://evitable.com/about-us" ], "scale": { "count": 5, "unit": "team members (employment basis unstated)", "bucket": "3-10", "safety_count": 5, "budget": "not published; fiscal sponsor Anti Entropy, EIN 88-0967420", "source": "https://evitable.com/about-us", "checked": "2026-08-31" }, "notes": "Canadian-LED more than Canadian-SCOPED: Krueger is Montréal-based (Mila) but Evitable's stated audience is the American public and it runs on a US fiscal sponsor. Krueger is also Senior Advisor on agentic AI at the Treasury Board of Canada Secretariat.", "editorial_scope": "core", "dedicated": false }, { "name": "Horizon Omega / Ω Labs / MASEG", "slug": "horizon-omega-labs-maseg", "website": "https://aisafetymontreal.org", "city": "Montréal", "province": "QC", "country": "Canada", "sector": "nonprofit", "kind": "community", "focus": [ "field-building" ], "status": "", "summary": { "en": "HΩ's own community work: Ω Labs paid coworking (opened May 2026), the MASEG public meetup, talks, AI Safety Papers We Love, and hackathons under the AI Safety Montréal banner.", "fr": "" }, "key_people": [ "Orpheus Lummis" ], "founded": "", "founded_year": "", "confidence": "AI verified", "last_verified": "", "sources": [ "https://www.meetup.com/montreal-ai-governance-ethics-safety/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "coi": "Horizon Omega maintains this directory. This entry describes the maintainer, and is written to the same standard and sourcing as every other.", "notes": "Listed for completeness: this is your org.", "editorial_scope": "core", "dedicated": false }, { "name": "IID (Institut intelligence et données, ULaval)", "slug": "iid-institut-intelligence-et-donnees-ulaval", "website": "https://iid.ulaval.ca", "city": "Québec City", "province": "QC", "country": "Canada", "sector": "university", "kind": "university-institute", "focus": [ "ethics-societal" ], "status": "", "summary": { "en": "Université Laval's AI and data institute, which \"fait progresser les savoirs et catalyse l'innovation en intelligence artificielle et en valorisation des données, au Québec et au-delà\" and pursues \"des recherches innovantes et éthiquement responsable en intelligence artificielle, science des données et toute autre discipline connexe\". Four research axes: physical environment; AI and data-processing methods; health and life sciences; and ethics. Responsible AI, ethics, privacy and cybersecurity — no alignment, evaluations, control or catastrophic-risk work.", "fr": "" }, "key_people": [ "Christian Gagné (Directeur)", "Lynda Robitaille (Directrice générale)", "Jean-François Lalonde (Directeur scientifique adjoint)", "Schallum Pierre (Conseiller sénior en IA responsable, éthique et innovation)" ], "founded": "", "founded_year": "", "confidence": "AI verified", "last_verified": "2026-08-31", "sources": [ "https://web.iid.ulaval.ca/a-propos" ], "scale": { "count": 100, "unit": "affiliated researchers", "bucket": "51-200", "safety_count": 0, "budget": "", "source": "https://web.iid.ulaval.ca/", "checked": "2026-08-31" }, "notes": "Own site fetched 2026-08-31; https://iid.ulaval.ca/ returns 200 and redirects to https://web.iid.ulaval.ca/. LOCATION is first-party from /nous-joindre: \"Université Laval, Pavillon Adrien Pouliot / 1065, avenue de la Médecine Local 1138, Québec (Québec) G1V 0A6.\" NO FRONTIER-RISK WORK, tested by counting over the full About text: \"alignement\" 0, \"existentiel\" 0, \"catastroph\" 0, \"sûreté\" 0, \"safety\" 0, \"risque\" 0. \"sécurité\" appears twice, both inside the ethics axis and both about privacy. Homepage \"sécurité\" hits are \"banc d'essai sécurisé et souverain\" and the \"ministère de la Cybersécurité et du Numérique\"; the one \"risque\" is deepfake risk in an electoral code of conduct. Safety people is therefore 0 by this map's definition, not \"not distinguished\". ⚠️ THIS CLOSES THE LAVAL QUESTION THE SAME WAY OBVIA DID: IID is the SECOND Laval AI vehicle and reaches the same verdict. Laval has real AI capacity and no safety lab. SCALE, verbatim first-party: \"Plus de 100 chercheurs et chercheuses membres\" — affiliated researchers with primary appointments, recorded as 100 for a \"plus de\" figure. Separately \"l'IID a déjà accompagné plus de 250 entreprises et organisations dans leur transition vers l'IA\", which is an ENGAGEMENT count and not people; it is recorded here and not in a scale field. ⚠️ FUNDING: a 2026-08-21 communiqué states \"Avec un financement de 2,1 M$ sur deux ans, l'expertise de l'Institut intelligence et données sera mobilisée pour mettre en place un banc d'essai sécurisé et souverain.\" That is a SINGLE PROJECT GRANT from Québec's Ministère de la Cybersécurité et du Numérique, NOT an operating budget, which is why the budget field is empty rather than carrying $2.1M. Founded: not published anywhere on the site.", "editorial_scope": "adjacent", "dedicated": false }, { "name": "IVADO", "slug": "ivado", "website": "https://ivado.ca", "city": "Montréal", "province": "QC", "country": "Canada", "sector": "consortium", "kind": "network", "focus": [ "ethics-societal" ], "status": "", "summary": { "en": "Québec's largest interdisciplinary AI research consortium, led by Université de Montréal with Polytechnique Montréal, HEC Montréal, Université Laval and McGill. Its framing phrase is \"robust, reasoning and responsible AI\" (the R3AI strategy), with research areas in ethics and responsibility and in explainability. Neither its home page nor its About page frames any stream as alignment, evaluations, control or catastrophic risk.", "fr": "" }, "key_people": [ "Nathalie de Marcellis-Warin (CEO, four-year mandate from 2026-09-01; succeeds Luc Vinet)", "Aaron Courville (Scientific Director)", "Christopher Pal (Interim Co-Scientific Director, Scientific and International Activities)", "Yoshua Bengio (Special Advisor and Founding Scientific Director)" ], "founded": "2016", "founded_year": "2016", "confidence": "AI verified", "last_verified": "2026-08-31", "sources": [ "https://ivado.ca/en/about-us/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "SCOPE CONFIRMED ADJACENT, first-party (ivado.ca, 2026-08-31): framing is R3AI (robust, reasoning and responsible AI); no stream is alignment, evaluations, control or catastrophic risk. Do not upgrade. FUNDING, recorded here rather than in the Budget field so it cannot be misread as safety money: the R3AI project won a $124.5M Canada First Research Excellence Fund grant in 2023, which IVADO says it will leverage into $500M over seven years. R3AI is trustworthy AI, NOT safety — by this map's definition IVADO's safety funding and safety headcount are both 0. SCALE: no researcher, member or staff count is published. Founded 2016 could not be confirmed first-party and is inherited from the HΩ landscape doc.", "editorial_scope": "adjacent", "dedicated": false }, { "name": "LawZero", "slug": "lawzero", "website": "https://lawzero.org", "city": "Montréal", "province": "QC", "country": "Canada", "sector": "nonprofit", "kind": "research-lab", "focus": [ "alignment-control" ], "status": "", "summary": { "en": "Building non-agentic \"Scientist AI\" to oversee agentic systems; a multi-step pathway to safe advanced AI. Incubated at Mila.", "fr": "" }, "key_people": [ "Yoshua Bengio" ], "founded": "2025-06", "founded_year": "2025", "confidence": "AI verified", "last_verified": "", "sources": [ "https://lawzero.org/en" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Funders per its own site: Gates Foundation, Future of Life Institute, Founders Pledge, Coefficient Giving, Schmidt Sciences, Silicon Valley Community Foundation, Jaan Tallinn. Feb 2026: federal letter of intent, reportedly >$100M under discussion; ~30 staff, plans to exceed 100. CORRECTION to the landscape doc, which had only $30M / Tallinn+Schmidt+OpenPhil+FLI. FIELD PROVENANCE: the 2025-06 founding date is from secondary sources, not lawzero.org.", "editorial_scope": "core", "dedicated": true }, { "name": "Mila: AI Safety Studio", "slug": "mila-ai-safety-studio", "website": "https://mila.quebec", "city": "Montréal", "province": "QC", "country": "Canada", "sector": "research institute", "kind": "funder-program", "focus": [ "alignment-control", "evaluations" ], "status": "", "summary": { "en": "Three stated priorities: guardrails (independent external control mechanisms on LLM apps), multi-turn and agentic safety benchmarking, and alignment research. First initiative = detecting/blocking harmful chatbot interactions with users in psychological distress.", "fr": "" }, "key_people": [ "Yoshua Bengio", "Hugo Larochelle", "David Krueger", "Simona Gandrabur" ], "founded": "2025-10", "founded_year": "2025", "confidence": "secondary", "last_verified": "", "sources": [ "https://mila.quebec/en/news/mila-launches-ai-safety-studio-with-first-initiative-to-help-prevent-suicide" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Launched 10 Oct 2025. ROOST partnership on open-source guardrails; national youth-safety hackathon Mar 2026. Krueger's group works on x-risk / alignment / interpretability. Mila also runs a biweekly safety seminar and a Community of Practice: AI Safety.", "editorial_scope": "core", "dedicated": true }, { "name": "Montreal AI Ethics Institute (MAIEI)", "slug": "montreal-ai-ethics-institute-maiei", "website": "https://montrealethics.ai", "city": "Montréal", "province": "QC", "country": "Canada", "sector": "nonprofit", "kind": "other", "focus": [ "ethics-societal" ], "status": "", "summary": { "en": "Democratising AI ethics literacy; publishes the State of AI Ethics Report.", "fr": "" }, "key_people": [ "Renjie Butalid (co-founder)", "Abhishek Gupta (co-founder, 1992–2024, in memoriam)" ], "founded": "", "founded_year": "", "confidence": "secondary", "last_verified": "", "sources": [ "https://aisecurityandsafety.org/en/organizations/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Ethics and literacy rather than catastrophic risk. FIELD PROVENANCE: description is a third-party directory paraphrase (aisecurityandsafety.org); montrealethics.ai not fetched.", "editorial_scope": "adjacent", "dedicated": false }, { "name": "NRC: AI safety & responsible AI", "slug": "nrc-ai-safety-responsible-ai", "website": "https://nrc.canada.ca/en/research-development/products-services/technical-advisory-services/ai-safety-responsible-ai", "city": "Ottawa", "province": "ON", "country": "Canada", "sector": "government", "kind": "government", "focus": [ "evaluations", "ethics-societal" ], "status": "", "summary": { "en": "Technical advisory services on AI safety and responsible AI; government benchmarking. CAISI partner.", "fr": "" }, "key_people": [], "founded": "", "founded_year": "", "confidence": "secondary", "last_verified": "", "sources": [ "https://nrc.canada.ca/en/research-development/products-services/technical-advisory-services/ai-safety-responsible-ai" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Assurance and compliance flavour rather than catastrophic risk, so kept adjacent.", "editorial_scope": "adjacent", "dedicated": true }, { "name": "OAISI (Ottawa AI Safety Initiative)", "slug": "oaisi-ottawa-ai-safety-initiative", "website": "https://ottawaaisafety.org/", "city": "Ottawa", "province": "ON", "country": "Canada", "sector": "student group", "kind": "student-group", "focus": [ "alignment-control", "evaluations", "interpretability", "security-robustness", "governance-policy", "field-building", "education" ], "status": "", "summary": { "en": "Technical research (alignment, interpretability, robustness, evaluation) and governance/policy upskilling: reading groups, speaker talks, workshops and monthly events, pointing members at MATS and ARENA. States its purpose as making AI \"human-aligned and beneficial\" by \"exploring the risks of advanced AI systems\".", "fr": "" }, "key_people": [], "founded": "", "founded_year": "", "confidence": "AI verified", "last_verified": "", "sources": [ "https://ottawaaisafety.org/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "FOUND 2026-08-28 outside the brief that prompted this pass, and it postdates the 2026-07-29 compilation: aisafety.com/communities added it 2026-08-09 (location Ottawa, Canada; activity Active; main focus AI safety). It is a uOttawa club, NOT Carleton - only uOttawa students can be members, though events are public. This amends the doc gap line \"uOttawa: no dedicated safety lab\", which is still true of a lab and now stale about the community. ACTIVITY SIGNALS CONFLICT, same shape as AI Alignment McGill: aisafety.com calls it Active, while its own site's events page reads \"No upcoming events are scheduled yet\" and the footer says (c) 2024. LEADERSHIP UNKNOWN: no organiser is named anywhere on the site; the only contact routes are a Google Form for uOttawa students and a mailing-list signup. Someone should join the mailing list or ask uOttawa faculty. FIELD PROVENANCE: everything above from ottawaaisafety.org itself except the activity level and listing date, which are from aisafety.com/communities; both fetched 2026-08-28.", "editorial_scope": "pipeline", "dedicated": true }, { "name": "OBVIA", "slug": "obvia", "website": "https://www.obvia.ca", "city": "Québec City", "province": "QC", "country": "Canada", "sector": "consortium", "kind": "network", "focus": [ "ethics-societal" ], "status": "", "summary": { "en": "Societal impacts of artificial intelligence and digital technologies: 290+ research members across 23 Québec post-secondary institutions and 7 research axes. Created at the initiative of the Fonds de recherche du Québec, which renewed its mandate in 2023 for a five-year period.", "fr": "" }, "key_people": [], "founded": "2019", "founded_year": "2019", "confidence": "AI verified", "last_verified": "2026-08-31", "sources": [ "https://www.obvia.ca/en/about-us" ], "scale": { "count": 290, "unit": "affiliated researchers (OBVIA: research members, 23 institutions)", "bucket": "200+", "safety_count": null, "budget": "", "source": "https://www.obvia.ca/en/about-us", "checked": "2026-08-31" }, "notes": "Societal-impact and ethics framing. Closes the doc's Laval question: this is Laval's AI vehicle, and it is not a safety lab.", "editorial_scope": "adjacent", "dedicated": false }, { "name": "PauseAI Canada", "slug": "pauseai-canada", "website": "https://pauseai.ca", "city": "", "province": "Country-wide", "country": "Canada", "sector": "nonprofit", "kind": "community", "focus": [ "governance-policy", "field-building" ], "status": "", "summary": { "en": "Canadian national chapter of the PauseAI movement: public advocacy for a pause on frontier AI development and for an international treaty, framed explicitly around existential and loss-of-control risk. Runs a bilingual EN/FR organizing platform with a stance questionnaire, a learning path and an \"Email your MP\" tool, plus a \"Warning Shot Protocol\" issue page arguing that Canada has no law requiring an independent safety assessment before a frontier AI system is built or deployed. Monthly Canada-wide video call and a continuous Montréal in-person programme: demonstrations, flyering, movie nights, coworking and Canadian-policy sessions.", "fr": "" }, "key_people": [ "Jeremy Eliosoff (principal organizer)", "\"nik l\" (Montréal co-organizer; no surname published)", "Olivier Gourment (author of the pauseai.ca platform)", "Étienne Langlois" ], "founded": "2025-11", "founded_year": "2025", "confidence": "AI verified", "last_verified": "2026-08-31", "sources": [ "https://pauseai.ca/en/about", "https://pauseai.info/communities" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "https://pauseai.ca/en/about", "checked": "2026-08-31" }, "coi": "Horizon Omega maintains this directory. Horizon Omega / Ω Labs events are cross-listed on the PauseAI Montréal Luma calendar, where Orpheus Lummis appears as host of five events and HΩ of three. Whoever maintains this entry is not a neutral outsider to it.", "notes": "Own site fetched 2026-08-31. NOT A LEGAL ENTITY: the About page calls it \"a developing Canadian group\" that \"collaborates with PauseAI Global\"; joining routes to PauseAI Global's own form; /act/sign and /act/actions redirect to pauseai.info; the privacy policy names no entity, address, incorporation or charity number, only info@pauseai.ca. The only legal entity in the movement's footer is Stichting PauseAI, kvk 92951031, a Dutch foundation. CHAPTER STATUS CONFIRMED BY THE PARENT: pauseai.info/communities enumerates 15 national groups and the Canada entry links pauseai.ca/en/. FOUNDED is the earliest first-party NATIONAL activity, not an incorporation: the monthly video call runs back to 2025-11-13. The Montréal chapter is older (\"Introducing PauseAI Montréal\", 2025-10-21) but that is the chapter's launch, not this Country-wide row's. ACTIVITY: 47 past Montréal events (2025-10-19 to 2026-08-30) plus 6 on the Canada-wide calendar and 2 upcoming; two Toronto meetups (2026-02-21, 2026-04-17) are the only non-Montréal in-person activity found. National in framing and infrastructure, Montréal-concentrated in practice — recorded here rather than in the City cell, which would misdescribe a national chapter and lose the Toronto thread. ATTENDANCE, from Luma RSVP counts and NOT membership: 10-24 per national call, 2-23 per Montréal event. No membership figure, team page, budget or funding statement is published anywhere, and pauseai.ca has no donation rail of its own — donations route to PauseAI Global. The site codebase is public (github.com/ogourment/pauseai.ca-website, created 2026-07-28), so the current site is one month old and under active development. SUPERSEDES the earlier \"CHECKED for Canadian chapters; none confirmed\" negative finding, which was wrong as to PauseAI; see the Encode AI entry, which is what remains of that row. Corroborated after the fact, not relied on: the AIGS Canada list carries \"Pause AI Canada\" with both \"Country-wide, Canada\" and \"Montréal, Canada\".", "editorial_scope": "core", "dedicated": false }, { "name": "Schwartz Reisman Institute (U of T)", "slug": "schwartz-reisman-institute-u-of-t", "website": "https://srinstitute.utoronto.ca", "city": "Toronto", "province": "ON", "country": "Canada", "sector": "university", "kind": "university-institute", "focus": [ "alignment-control", "interpretability", "governance-policy" ], "status": "", "summary": { "en": "Interpretability, alignment and oversight research alongside technology governance. Named Grosse and Duvenaud as SR Chairs.", "fr": "" }, "key_people": [ "David Lie (Director)", "Roger Grosse", "David Duvenaud", "Sheila McIlraith", "Karina Vold" ], "founded": "", "founded_year": "", "confidence": "unverified (doc)", "last_verified": "", "sources": [], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Founding year not verified this pass.", "editorial_scope": "core", "dedicated": true }, { "name": "SecuritAI", "slug": "securitai", "website": "https://securitai.io", "city": "Toronto", "province": "ON", "country": "Canada", "sector": "company", "kind": "company", "focus": [ "evaluations", "security-robustness" ], "status": "", "summary": { "en": "Adversarial red-teaming of LLMs, chatbots and agents plus a runtime \"AI firewall\".", "fr": "" }, "key_people": [], "founded": "", "founded_year": "", "confidence": "secondary", "last_verified": "", "sources": [ "https://securitai.io/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Security product, not safety research.", "editorial_scope": "adjacent", "dedicated": true }, { "name": "SFU Trustworthy AI Lab (TAI Lab)", "slug": "sfu-trustworthy-ai-lab-tai-lab", "website": "https://sfu-tai.github.io", "city": "Burnaby", "province": "BC", "country": "Canada", "sector": "university", "kind": "research-lab", "focus": [ "security-robustness", "ethics-societal" ], "status": "", "summary": { "en": "Certifiable and verifiable trustworthiness guarantees (robustness, fairness, numerical reliability) for large-scale deep learning and foundation models.", "fr": "" }, "key_people": [ "Linyi Li (Director)" ], "founded": "2024 (lab)", "founded_year": "2024", "confidence": "secondary", "last_verified": "", "sources": [ "https://sfu-tai.github.io/research" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Li joined SFU in 2024. BC's main dedicated technical group. The doc flagged BC coverage as thin and this is what fills it.", "editorial_scope": "core", "dedicated": false }, { "name": "Starling Centre for Just Technologies and Just Societies (Western)", "slug": "starling-centre-for-just-technologies-and-just-societies-western", "website": "https://starlingcentre.ca", "city": "London", "province": "ON", "country": "Canada", "sector": "university", "kind": "university-institute", "focus": [ "governance-policy", "ethics-societal" ], "status": "", "summary": { "en": "Critical data-studies centre in the Faculty of Information and Media Studies at Western University, working \"toward more just, equitable and humane futures for all\" by \"assessing the impacts of digital data, artificial intelligence, and related technologies\". Its AI work is oversight of deployed systems: \"strengthen AI oversight in Canada through empirical investigations of the impacts of automation... contributions to civil society consultations and co-developing policy interventions\". Projects include AI Governance in Canada, Situating London's AI Homelessness Model, Insights and Interventions for AI Policy, Toward a Canadian Right to Repair, and Preventing Data Harms in Canada.", "fr": "" }, "key_people": [ "Co-Directors: Alissa Centivany", "Alison Hearn", "Joanna Redden", "Luke Stark (no single director)" ], "founded": "", "founded_year": "", "confidence": "secondary", "last_verified": "2026-08-31", "sources": [ "https://starlingcentre.ca/about/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": 0, "budget": "", "source": "https://starlingcentre.ca/research-team/", "checked": "2026-08-31" }, "notes": "Own site fetched 2026-08-31: /, /about/, /research-team/, /projects/, /contact/ and /?s=AI+safety, all HTTP 200. ⚠️ FIELD PROVENANCE — city and province come from Western University's own site via the Centre's stated faculty affiliation (\"Located in the Faculty of Information and Media Studies at Western University\"); starlingcentre.ca publishes NO address or city on any page, including /contact/. That is why confidence is secondary rather than verified: the load-bearing location field is not first-party to the Centre. Do NOT take the \"City of London, Ontario\" that appears in their CHAI project text — that is their research subject, not their address. AIGS lists \"London, Canada\", which agrees but is third party and is not what this row rests on. NO AI SAFETY WORK, and the negative was tested by counting over the About and Projects text: \"AI safety\" 0, \"alignment\" 0, \"existential\" 0, \"catastroph\" 0, \"extinction\" 0, \"misalign\" 0. The single \"frontier\" hit is the funder name \"New Frontiers in Research Fund\". Their own site search for \"AI safety\" returns two critical-social-science talks. Safety people is therefore recorded as 0, not \"not distinguished\". WHY CARRY IT: a named university centre doing Canadian AI-governance research, and Western is otherwise absent from this map. SCALE: no first-party figure — the research-team page lists bios but states no count. FUNDING is disclosed only per project and without amounts: \"Initially funded by a 2023 New Frontiers in Research Fund - Exploration grant\" and \"funded by FIMS, Western University\". Founded: not published.", "editorial_scope": "adjacent", "dedicated": false }, { "name": "TAISI (Toronto AI Safety Initiative)", "slug": "taisi-toronto-ai-safety-initiative", "website": "https://www.taisi.ca", "city": "Toronto", "province": "ON", "country": "Canada", "sector": "student group", "kind": "student-group", "focus": [ "alignment-control", "governance-policy", "education" ], "status": "", "summary": { "en": "Trains people to become AI safety researchers via a fellowship and a summer intensive; technical and policy work on long-term/catastrophic risk from advanced AI.", "fr": "" }, "key_people": [ "Joseph Kostousov" ], "founded": "2025", "founded_year": "2025", "confidence": "AI verified", "last_verified": "", "sources": [ "https://www.taisi.ca/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Founded at the start of the 2025-26 academic year; Kairos-seeded. Registered on U of T's Student Organization Portal.", "editorial_scope": "pipeline", "dedicated": true }, { "name": "The AI Safety Foundation (AISF)", "slug": "the-ai-safety-foundation-aisf", "website": "https://www.aisfoundation.ai", "city": "Toronto", "province": "ON", "country": "Canada", "sector": "nonprofit", "kind": "other", "focus": [ "alignment-control", "education" ], "status": "", "summary": { "en": "Increases awareness and scientific understanding of the catastrophic risks of AI through education and research. Runs the annual Hinton Lectures and the CJF-Hinton Award for Excellence in AI Safety Reporting. Event-driven.", "fr": "" }, "key_people": [ "Geoffrey Hinton (co-founder", "board member)" ], "founded": "", "founded_year": "", "confidence": "secondary", "last_verified": "", "sources": [ "https://www.aisfoundation.ai/" ], "scale": { "count": 2, "unit": "FTE staff (apparent)", "bucket": "1-2", "safety_count": null, "budget": "not published; registered Canadian charity BN 78867 2822 RR0001", "source": "https://www.aisfoundation.ai/about", "checked": "2026-08-31" }, "notes": "CJF-Hinton Award launched Oct 2025 with the Canadian Journalism Foundation. 2025 Hinton Lectures: 10-12 Nov, Bassett Theatre, Toronto. Founding year not verified.", "editorial_scope": "core", "dedicated": true }, { "name": "The Dais (TMU)", "slug": "the-dais-tmu", "website": "https://dais.ca", "city": "Toronto", "province": "ON", "country": "Canada", "sector": "university", "kind": "think-tank", "focus": [ "governance-policy" ], "status": "", "summary": { "en": "Public policy and leadership think tank at Toronto Metropolitan University, in its own words \"working at the intersection of technology, education and democracy\", across three programme areas: Innovation and Tech; Education and Skills; Democracy and Trust. Its AI policy work is adoption, skills, compute and governance — self-reported as \"informed by over 25 reports\" — and includes a submission to the consultation on Canada's renewed AI Strategy (October 2025). No dedicated AI safety work; its \"safety\" usage is consistently online safety, child protection and deepfakes.", "fr": "" }, "key_people": [ "André Côté (Interim Executive Director)", "Sam Andrey", "Christelle Tessono" ], "founded": "", "founded_year": "", "confidence": "AI verified", "last_verified": "2026-08-31", "sources": [ "https://dais.ca/about/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": 0, "budget": "", "source": "https://dais.ca/about/", "checked": "2026-08-31" }, "notes": "Own site fetched 2026-08-31: dais.ca/, /about/, /contact/, several site searches, and their own PDF submission. LOCATION is clean and first-party, from /contact/: \"20 Dundas Street West Suite 921 Toronto, ON M5G 2C2\", dais@torontomu.ca. STRONG NEGATIVE ON SAFETY, AND IT WAS TESTED HARD: \"AI safety\", \"alignment\", \"existential\", \"catastrophic\", \"frontier\", \"extinction\" and \"misalign\" do not appear in the About page text; the only sitewide \"existential\" hit is \"the existential threat of climate change\". Their flagship AI-policy output, the October 2025 submission to Canada's renewed AI strategy, was downloaded and counted: existential 0, catastrophic 0, frontier 0, extinction 0, misuse 0, systemic risk 0, safety institute 0, red team 0. Its single \"AI safety\" occurrence describes INDUSTRY RHETORIC: \"The primary narrative advanced by large AI firms... is that AI safety and regulation stifle innovation.\" Safety people is therefore 0, not \"not distinguished\". The nearest safety touchpoint is a press mention rather than a programme: Christelle Tessono commenting on the efficacy of Canada's new AI Safety Institute. A staffer quoted in media is not a basis for core. ⚠️ FOUNDED — DO NOT RESOLVE TO 2015. The About page gives lineage only: \"Founded in 2015, the Brookfield Institute became...\", \"The Leadership Lab was created in 2017...\", \"In 2023, ... the two organizations combined our talents.\" The Dais itself has no published founding year; 2023 is the defensible merger date if one is ever needed, and must be flagged as such. SCALE: no staff or budget figure for the Dais. ⚠️ One historical PREDECESSOR figure exists and must not be attached to this row: \"The Brookfield Institute was made possible through a generous $16 million donation from Jack Cockwell along with the Brookfield Partners Foundation.\" METHOD CAVEAT: dais.ca/?s= is WordPress relevance-ranked, not exact-match, so site-search results are indicative only; the hard negatives above are grep counts over text actually downloaded.", "editorial_scope": "adjacent", "dedicated": false }, { "name": "Trajectory Labs", "slug": "trajectory-labs", "website": "https://www.trajectorylabs.org", "city": "Toronto", "province": "ON", "country": "Canada", "sector": "nonprofit", "kind": "hub", "focus": [ "field-building" ], "status": "", "summary": { "en": "AI safety coworking space in downtown Toronto for 30+ practitioners, weekly events from casual meetups to seminars and weekend hackathons, plus a peer/mentorship network. Open Philanthropy funded.", "fr": "" }, "key_people": [ "Mario Gibney (Founder/ED)", "Joanne Kim (COO)", "Georgia Berg (Ops)", "board: Giles Edkins", "Kathrin Gardhouse" ], "founded": "2025-01", "founded_year": "2025", "confidence": "AI verified", "last_verified": "", "sources": [ "https://www.trajectorylabs.org/about-us" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Grew out of the Toronto AI Safety / \"Toronto AI Aligners\" meetup, which it now hosts. Leadership has CHANGED since the landscape doc (which listed Juliana Eberschlag on ops). NAME TRAP: trajectorylabs.com (a US evals PBC) and trajectoryai.org are unrelated orgs.", "editorial_scope": "core", "dedicated": false }, { "name": "TrojAI", "slug": "trojai", "website": "https://troj.ai", "city": "Saint John", "province": "NB", "country": "Canada (US-owned)", "sector": "company", "kind": "company", "focus": [ "security-robustness" ], "status": "", "summary": { "en": "AI model and agent security platform: detects backdoors, trojans and data poisoning pre-deployment; \"Defend for MCP\" for agent workflows.", "fr": "" }, "key_people": [ "James Stewart", "Stephen Goddard (founders)" ], "founded": "2019", "founded_year": "2019", "confidence": "secondary", "last_verified": "", "sources": [ "https://troj.ai/company/about-us/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "RESOLVED 2026-07-29: acquisition confirmed, announced 15 Jun 2026 by A10 Networks (Silicon Valley), price undisclosed. Now a US-owned subsidiary, but the 22-person team stays in Saint John and both founders state New Brunswick operations continue and are expected to grow. So: Canadian operations, US ownership. HQ Saint John NB with a Boston office; the only Atlantic-Canada AI-security company found.", "editorial_scope": "adjacent", "dedicated": false }, { "name": "UBC AI Safety", "slug": "ubc-ai-safety", "website": "https://www.ubcaisafety.org", "city": "Vancouver", "province": "BC", "country": "Canada", "sector": "student group", "kind": "student-group", "focus": [ "alignment-control", "governance-policy", "field-building", "education" ], "status": "", "summary": { "en": "AI safety education and community at UBC: technical intro course, ML accelerator, technical paper reading group, policy reading group, and an AI Policy Review. Explicitly framed around alignment and catastrophic outcomes.", "fr": "" }, "key_people": [ "Student exec team (no named lead on site)" ], "founded": "", "founded_year": "", "confidence": "AI verified", "last_verified": "", "sources": [ "https://www.ubcaisafety.org/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Contact plus a Slack community. Confirms a real UBC pipeline presence, which earlier searches doubted.", "editorial_scope": "pipeline", "dedicated": true }, { "name": "UBC CAIDA", "slug": "ubc-caida", "website": "https://caida.ubc.ca", "city": "Vancouver", "province": "BC", "country": "Canada", "sector": "university", "kind": "university-institute", "focus": [ "ethics-societal" ], "status": "", "summary": { "en": "ICICS Centre for AI Decision-making and Action. Its own site does not mention AI safety or trustworthy AI; recorded here as adjacent context.", "fr": "" }, "key_people": [], "founded": "", "founded_year": "", "confidence": "secondary", "last_verified": "", "sources": [ "https://caida.ubc.ca/caida-ubc-icics-centre-artificial-intelligence-decision-making-and-action" ], "scale": { "count": 100, "unit": "faculty (network affiliation across 30 UBC departments)", "bucket": "51-200", "safety_count": null, "budget": "", "source": "https://caida.ubc.ca/", "checked": "2026-08-31" }, "notes": "Not a safety centre. UBC's actual safety links are the CAISI research council (Mathias Lécuyer) and the student group.", "editorial_scope": "adjacent", "dedicated": false }, { "name": "UVic AI (UVicAI)", "slug": "uvic-ai-uvicai", "website": "https://uvicai.ca", "city": "Victoria", "province": "BC", "country": "Canada", "sector": "student group", "kind": "student-group", "focus": [ "field-building", "education" ], "status": "", "summary": { "en": "AI safety work is DORMANT. From summer 2023 to June 2024 the club ran an \"AI Ethics Reading Group\" that was, in its own words, \"focusing on AI safety research, including value alignment, interpretability, AI governance, and societal impact\" — 42 numbered sessions covering Anthropic's Core Views on AI Safety, deceptive alignment, mechanistic interpretability, Coherent Extrapolated Volition, goal misgeneralisation, OpenAI's Superalignment, agent foundations, intelligence explosions, and 80,000 Hours' Preventing an AI-related catastrophe. That group has stopped. The club itself continues on general AI: ML projects, competitions, workshops and guest talks.", "fr": "" }, "key_people": [], "founded": "", "founded_year": "", "confidence": "AI verified", "last_verified": "2026-08-31", "sources": [ "https://uvicai.ca/2024/06/18/AI-Ethics-Reading-Group.html" ], "scale": { "count": 655, "unit": "community members", "bucket": "200+", "safety_count": 0, "budget": "", "source": "https://discord.com/api/v9/invites/zbBNT8Spjf?with_counts=true", "checked": "2026-08-31" }, "notes": "Own site fetched 2026-08-31. THE READING GROUP PAGE STATES VERBATIM: \"Note: This reading group does no longer run. With sufficient interest we can bring this back. Feel free to email us if you want this to come back.\" Last session listed is Session 42, Mon Jun 3 (2024); Session 1 was 2023-05-23. Despite the \"AI Ethics\" name the curriculum is unambiguously AI safety, not AI ethics. The separate Summer 2024 schedule sheet corroborates it (Bostrom's Superintelligence ch. 4, Transformer Circuits) and shows the decay: rows are filled through 2024-06-24 and blank for every scheduled date from 2024-07-01 to 2024-08-12. THE CURRENT SITE CARRIES NO AI SAFETY FRAMING: About says only \"We are a community of students at the University of Victoria exploring artificial intelligence from both technical and ethical perspectives\", and \"AI safety\", \"alignment\" and \"existential\" return zero hits across /, /about/ and /projects/. The reading-group page is reachable only via the sitemap, not the navigation. THE PARENT CLUB IS ALIVE AND HAS MOVED ON: © 2026 footer; project posts are 3 in 2025, 2 in 2024, 1 in 2023, and the 2025 output is Sakura bloom trends, chemical toxicology with ML, and learning resources. No safety content since the reading group stopped. ⚠️ THIRD-PARTY CLAIM NOT INHERITED: aisafety.com/communities records this group as \"Very active\", \"Partial focus on AI safety\", size 450; the activity level is about the CLUB, not about safety work, and neither it nor the 450 is used here. ⚠️ THREE SCALE FIGURES EXIST AND THEY DISAGREE, all recorded rather than reconciled: 655 Discord members with 64 online from Discord's public UNAUTHENTICATED invite endpoint on 2026-08-31 (no server joined) — the figure in the scale field; 661 on the club's own Linktree the same day; and 450 per aisafety.com. This counts Discord members of a general AI club, NOT AI safety participants and NOT club membership. Safety people is 0 because the safety programme is stopped, not because it is undistinguished. No key people are published on the site, the about page, the Linktree or GitHub. No funding published. ⚠️ SCOPE IS CONTESTED: verify-candidates.md says pipeline (a revivable 42-session alignment group with a standing public offer to restart), gaps-and-discovery.md says excluded or \"pipeline (dormant)\" and \"do not record it as an active AI safety group\". Drafted as pipeline; the facts do not change either way. WHY IT MATTERS: the only Vancouver Island entry in the dataset, and a restart candidate rather than a cold lead.", "editorial_scope": "pipeline", "dedicated": false }, { "name": "Vancouver AI Safety Hub (VASH)", "slug": "vancouver-ai-safety-hub-vash", "website": "https://vancouveraisafety.org", "city": "Vancouver", "province": "BC", "country": "Canada", "sector": "nonprofit", "kind": "community", "focus": [ "field-building", "alignment-control", "governance-policy" ], "status": "", "summary": { "en": "Field-building for AI safety in Metro Vancouver: connecting people working on technical safety, governance, and comms and advocacy. Runs coworking sessions, a strategy round table and project spotlights, with working groups and applied sprints planned.", "fr": "" }, "key_people": [ "Krystal Pan (co-founder)", "Sana Shams (co-founder)", "Ibrahim El-Chami (co-founder)" ], "founded": "2026", "founded_year": "2026", "confidence": "AI verified", "last_verified": "2026-08-31", "sources": [ "https://vancouveraisafety.org" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "https://vancouveraisafety.org", "checked": "2026-08-31" }, "notes": "Own site fetched 2026-08-31. Not on the AIGS Canada list and not previously here. SCOPE: core on framing, pipeline on track record. Its stated object is AI safety including alignment research, but it is roughly three months old: earliest dated programme July 2026, and the August 2026 round table is described as its first. No research output, no published funding, no legal form stated (sector inferred). RELATIONSHIP: co-founder Sana Shams has helped lead UBC AI Safety since 2024, so this and the UBC AI Safety row are not independent capacity. Her bio also states she is an AIGS Canada Vancouver chapter organizer, which means AIGS has a Vancouver chapter our Ottawa-only AIGS row does not reflect; confirm with AIGS.", "editorial_scope": "pipeline", "dedicated": true }, { "name": "Vector Institute: AI trust & safety", "slug": "vector-institute-ai-trust-safety", "website": "https://vectorinstitute.ai", "city": "Toronto", "province": "ON", "country": "Canada", "sector": "research institute", "kind": "university-institute", "focus": [ "alignment-control", "evaluations", "security-robustness", "governance-policy", "ethics-societal" ], "status": "", "summary": { "en": "Independent frontier-model evaluations (2025: benchmarked 11 LLMs), adversarial robustness, trustworthy ML, trust/safety policy. Weighted toward evals and governance rather than deep alignment.", "fr": "" }, "key_people": [ "Roger Grosse (Faculty Member; also at Anthropic)", "David Duvenaud (founding member)" ], "founded": "2017 (institute)", "founded_year": "2017", "confidence": "unverified (doc)", "last_verified": "", "sources": [ "https://aisecurityandsafety.org/en/organizations/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "One of Canada's three national AI institutes; CAISI partner.", "editorial_scope": "core", "dedicated": true }, { "name": "Waterloo Cybersecurity & Privacy Institute", "slug": "waterloo-cybersecurity-privacy-institute", "website": "https://uwaterloo.ca/cybersecurity-privacy-institute/", "city": "Waterloo", "province": "ON", "country": "Canada", "sector": "university", "kind": "university-institute", "focus": [ "governance-policy" ], "status": "", "summary": { "en": "Cybersecurity and privacy research institute. Its own site does not describe AI safety work; recorded here as adjacent context.", "fr": "" }, "key_people": [], "founded": "", "founded_year": "", "confidence": "secondary", "last_verified": "", "sources": [], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "", "editorial_scope": "adjacent", "dedicated": false } ], "checked_not_listed": [ { "name": "AI4Society (University of Alberta) — declares itself not active, and no AI safety work (checked 2026-08-31)", "slug": "ai4society-university-of-alberta-declares-itself-not-active-and-no-ai-safety-work-checked-2026-08-31", "website": "https://ai4society.ca", "city": "Edmonton", "province": "AB", "country": "Canada", "sector": "university", "kind": "", "focus": [], "status": "", "summary": { "en": "Nothing. AI4Society was one of five Signature Areas of Research and Teaching at the University of Alberta, \"focused on Artificial Intelligence, its applications, and its transformative role for our society\". It declares itself inactive on its own homepage, and no AI safety work appears anywhere on it.", "fr": "" }, "key_people": [ "Dr. Eleni Stroulia (Director)", "Dr. Geoffrey Rockwell (Associate Director)", "Dr. Nicolás Arnáez (Project Coordinator) — all as published, undated" ], "founded": "", "founded_year": "", "confidence": "negative finding", "last_verified": "2026-08-31", "sources": [ "https://ai4society.ca/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "CHECKED 2026-08-31. TWO INDEPENDENT GROUNDS TO EXCLUDE. (1) THE ORGANIZATION DECLARES ITSELF DORMANT, displayed twice on its own homepage, verbatim: \"Please note that AI4Society is not active at this moment, we appreciate your understanding\". Footer copyright reads © 2022; the newest featured content is 2022; /contact-us/ returns their own 404. (2) NO AI SAFETY WORK, CATEGORICALLY — counts over the full homepage text: \"safety\" 0, \"alignment\" 0, \"existential\" 0, \"catastrophic\" 0, \"frontier\" 0, \"risk\" 0. Not one occurrence of any of them. URL NOTE: https://www.ualberta.ca/ai4society/ returns HTTP 403 to a plain client and, with a browser user-agent, redirects to a ualberta.ca path that returns HTTP 404. The surviving site is ai4society.ca, which is also the URL the AIGS Canada list carries, so the dead university URL is not their error. Location from the site footer: \"AI4Society - VPRI Office / Address: Faculty of Science, University of Alberta / 6-211 CCIS | Edmonton, AB | T6G 2E1.\" SCALE, IF EVER NEEDED, IS A 2022 SNAPSHOT AND MUST BE DATED AS ONE: \"225 Researchers – 11 Faculties – 87 International institutes – 45 projects – 65 events\", on a page stamped © 2022 with no content newer than 2022. A third-party summary claimed 133 researchers; it is not used, and the disagreement is itself the reason to date the figure. No scale fields are populated on this entry. Alberta stays represented by Amii. ⚠️ READ-ACROSS: this is the clearest single instance of the staleness measured in prior-art.md §1.3 — a directory row can be four years old and point at an organization that has since shut down.", "editorial_scope": "excluded", "dedicated": false }, { "name": "ALIGN the World — no evidence of activity since August 2023 (checked 2026-08-31)", "slug": "align-the-world-no-evidence-of-activity-since-august-2023-checked-2026-08-31", "website": "https://aligntheworld.com", "city": "", "province": "", "country": "Canada", "sector": "nonprofit", "kind": "", "focus": [], "status": "", "summary": { "en": "Physician-founded not-for-profit doing public advocacy on AI risk, with explicit catastrophic framing: \"unacceptable safety risks associated with the AI industry as it operates now, including the possible mass loss of human life.\" Two-point mandate: \"1. To raise awareness about the potential dangers of AI systems. 2. To put pressure on governments and AI companies to regulate this industry responsibly.\" The name is an acronym — Alliance for the Long-term Integration, Governance, and Nurturance of the World. No evidence of any activity after August 2023.", "fr": "" }, "key_people": [ "Matias Raski (director; identified only via the Twitter link labelled \"ALIGN's director's Twitter page\" on their About page)" ], "founded": "", "founded_year": "", "confidence": "negative finding", "last_verified": "2026-08-31", "sources": [ "https://aligntheworld.com/about-us/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "Own site fetched 2026-08-31. FIELD PROVENANCE: the mandate and self-description are verified first-party from aligntheworld.com; confidence is \"negative finding\" because the load-bearing result of this check is the ABSENCE of activity, and because city, province and country have no first-party source at all. DORMANT — NO EVIDENCE OF ANY ACTIVITY AFTER AUGUST 2023. The homepage still advertises as UPCOMING an \"AI Safety Rally in Vancouver, Canada — Coming up on August 26th, 2023 at the Vancouver Art Gallery\", three years after the date. The events page lists only three May-2023 items whose location string reads \"Canda\" [sic] and which appear to be unreplaced WordPress theme demo content, not real events. The blog page renders a heading and no posts. No YouTube video listing could be retrieved from @aligntheworld (2026-08-31). No team page, no staff names beyond the director's Twitter handle, no financials. The About page states intent, not fact — \"operating within the legal constraints of a registered not-for-profit organization\" — with NO jurisdiction and NO registration number given. ⚠️ LOCATION NOT INHERITED: the site never states a city, province or country for the organization. Its only geographic tie is that a 2023 rally was held in Vancouver, so Country is an inference from that plus the AIGS listing, not a first-party fact. The AIGS Canada list carries both \"Vancouver, Canada\" and \"Country-wide, Canada\"; neither is inherited here. ⚠️ SCOPE IS CONTESTED (§3.4.2): the task brief directs a negative-finding entry; verify-candidates.md leaned pipeline because ALIGN names a real person, states a coherent x-risk mandate and did hold a real public event. Written so that flipping scope to pipeline changes one field and nothing else. Re-check before any outreach.", "editorial_scope": "excluded", "dedicated": false }, { "name": "AVERI (AI Verification and Evaluation Research Institute) — a US 501(c)(3), not a Canadian organization (checked 2026-08-31)", "slug": "averi-ai-verification-and-evaluation-research-institute-a-us-501-c-3-not-a-canadian-organization-checked-2026-08-31", "website": "https://www.averi.org", "city": "", "province": "", "country": "United States", "sector": "nonprofit", "kind": "", "focus": [], "status": "", "summary": { "en": "Core AI safety by this map's definition — third-party auditing of frontier AI developers' safety and security claims, pilot audits with frontier labs, auditing standards (co-author of AEF-1 via the AI Evaluator Forum), and audit-requirement policy advocacy (endorsed the legislation that became Illinois SB 315, the first US-based frontier AI audit requirement). Mission in its own words: \"to make frontier AI auditing effective and universal\". Excluded on geography only.", "fr": "" }, "key_people": [ "Miles Brundage (Executive Director; ex-OpenAI Policy Research and AGI Readiness, ex-FHI)", "Sean McGregor (Lead Research Engineer; founder of the AI Incident Database)", "Carly Tryens (Executive Operations)", "Jacqueline Grimsley (Operations Director)" ], "founded": "", "founded_year": "", "confidence": "negative finding", "last_verified": "2026-08-31", "sources": [ "https://www.averi.org/about" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "CHECKED 2026-08-31, not Canadian. Their own site footer, on every page, verbatim: \"The AI Verification and Evaluation Research Institute (AVERI, which rhymes with bravery) is a 501(c)(3) US-based non-profit organization.\" Careers page: \"AVERI is remote-first and globally distributed.\" No Canadian office, entity, programme or staff member is published anywhere on the site, and none of the four named people is published as Canada-based. Surfaced as a candidate only because one AVERI role carries the 80,000 Hours \"AI safety\" tag — the tag is right about the work and says nothing about the country. Recorded so nobody researches it twice. ⚠️ CROSS-REFERENCE, useful and not a reason to list them: Lead Research Engineer Sean McGregor founded the AI Incident Database, which is directly relevant to HΩ's CAIM/FAIM work on who collects AI incidents in Canada.", "editorial_scope": "excluded", "dedicated": false }, { "name": "Carleton AI safety presence", "slug": "carleton-ai-safety-presence", "website": "", "city": "Ottawa", "province": "ON", "country": "Canada", "sector": "", "kind": "", "focus": [], "status": "", "summary": { "en": "No dedicated AI safety group, lab or centre at Carleton, and no AI safety student group. What exists: the Carleton AI Society, a CUSA club covering data science, AI and ML broadly with an ethics-and-social-implications framing; the Trustworthy AI Lab in the Technology Innovation Management program (Z-Inspection trustworthiness assessments, responsible-AI adoption and education); the AIML lab (general AI/ML); and TrustCAV, an NSERC CREATE training program on connected and autonomous vehicles across five universities, where safety is vehicle safety. None addresses alignment or catastrophic risk. Carleton's own AI hub page names no safety entity at all.", "fr": "" }, "key_people": [], "founded": "", "founded_year": "", "confidence": "negative finding", "last_verified": "", "sources": [ "https://carleton.ca/ai/ + carleton.ca/tim/trustworthy-ai-lab/ + cusaclubs.ca/club/carleton-ai-society (fetched 2026-08-28)" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "CHECKED 2026-08-28, nothing dedicated found. Closes a gap the landscape doc never opened: its \"Gaps / to verify\" section named uOttawa, Queen's, Concordia and Laval and never checked Carleton. The individual capacity that does exist at Carleton has its own rows (Carleton SCS: CRC in AI in the core block; Carleton SPPA: CRC in Governance and AI in the adjacent block); this row records that there is no group, lab or centre above them. aisafety.com/communities lists no Carleton entry. Recorded so the gap isn't re-searched from scratch.", "editorial_scope": "excluded", "dedicated": false }, { "name": "Concordia AI", "slug": "concordia-ai", "website": "https://concordia-ai.com", "city": "Beijing", "province": "", "country": "China", "sector": "company", "kind": "", "focus": [], "status": "", "summary": { "en": "AI safety and governance work bridging Chinese and international experts; advises multilateral organisations.", "fr": "" }, "key_people": [ "Brian Tse" ], "founded": "", "founded_year": "", "confidence": "secondary", "last_verified": "", "sources": [ "https://concordia-ai.com/about-us/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "⚠️ NAME TRAP: nothing to do with Concordia University in Montréal. Beijing-based, so out of scope. Recorded to stop it being mis-added later.", "editorial_scope": "excluded", "dedicated": false }, { "name": "Encode AI", "slug": "encode-ai", "website": "https://encodeai.org", "city": "", "province": "", "country": "International", "sector": "nonprofit", "kind": "", "focus": [], "status": "", "summary": { "en": "US youth-led AI policy and legislative advocacy. No Canadian entity confirmed.", "fr": "" }, "key_people": [], "founded": "", "founded_year": "", "confidence": "negative finding", "last_verified": "", "sources": [], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "US youth-led AI policy organization. ⚠️ THIS PASS DID NOT RE-CHECK ENCODE. The 2026-07-29 finding — \"checked for Canadian chapters; none confirmed\" — stands unchanged and unverified as of 2026-09-01; its only source is that search, so no source URL and no last_verified date are recorded here. This entry was split out of the former composite \"PauseAI / Encode AI\" row on 2026-09-01, when the PauseAI half was falsified by first-party evidence and became its own entry, PauseAI Canada. DO NOT READ THE PAUSEAI CORRECTION AS APPLYING HERE: PauseAI Canada is a live, active, national volunteer chapter; nothing of the kind has been established or disproved for Encode, because nobody looked. The original row's own caution still applies to Encode specifically: grassroots chapters are easy to miss, so this is worth re-checking with the network.", "editorial_scope": "excluded", "dedicated": false }, { "name": "Factor Omega", "slug": "factor-omega", "website": "https://factoromega.org", "city": "", "province": "", "country": "USA", "sector": "nonprofit", "kind": "", "focus": [], "status": "", "summary": { "en": "Nonprofit resource hub redeploying strategic resources such as GPUs, plus advocacy and consultancy for AI safety projects.", "fr": "" }, "key_people": [ "Saturnin Pugnet (Founder/President)", "Matthieu Kratz", "Karim Beguir", "Pablo Eder" ], "founded": "", "founded_year": "", "confidence": "AI verified", "last_verified": "", "sources": [ "https://factoromega.org/" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "⚠️ CORRECTION to the HΩ landscape doc, which listed it as \"Canada (HQ unconfirmed)\". Its own site states a US 501(c)(3). NOT a Canadian org.", "editorial_scope": "excluded", "dedicated": false }, { "name": "Halifax LessWrong Meetup — a general rationality meetup, not an AI safety group (checked 2026-08-31)", "slug": "halifax-lesswrong-meetup-a-general-rationality-meetup-not-an-ai-safety-group-checked-2026-08-31", "website": "https://www.lesswrong.com/groups/fxYRzdApLuZ8cTcfR", "city": "Halifax", "province": "NS", "country": "Canada", "sector": "community", "kind": "", "focus": [], "status": "", "summary": { "en": "Not an AI safety group. Its own one-line self-description is \"A meetup group for rationalists in Halifax, NS.\" Two of ten listed sessions were AI-risk topics — AI Timelines & Strategy (Jan 2025) and Open-Source AI and Bioterrorism Risk (Nov 2023); the rest are prediction markets, institution design, Abundance, Simulators, and similar general-rationality material.", "fr": "" }, "key_people": [ "Organizers published only as LessWrong handles: interstice, Ideopunk" ], "founded": "", "founded_year": "", "confidence": "negative finding", "last_verified": "2026-08-31", "sources": [ "https://www.lesswrong.com/groups/fxYRzdApLuZ8cTcfR" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "CHECKED 2026-08-31 (the page is behind a Vercel challenge to plain curl; retrieved via WebFetch). A general rationality meetup, not an AI safety group. Contact discord.gg/kXFaGQBB5h. Nine past events listed plus one upcoming: Simulators (Mar 15) · The Electric Slide (Jan 18) · The Sort (Nov 9, 2025) · Abundance (Jun 22, 2025) · Amish NWO (Jun 8, 2025) · AI Timelines & Strategy (Jan 26, 2025) · Institution Design and Coordination (Oct 13, 2024) · Halifax Rationality Meetup (Feb 18, 2024) · Open-Source AI and Bioterrorism Risk (Nov 19, 2023) · Prediction Markets (Aug 12, 2023). A \"Load More\" control indicates further history. It publishes no website beyond the LessWrong group page, has no named organizers beyond handles, and has no organizational form. ⚠️ METHOD NOTE THAT IS LOAD-BEARING: LessWrong's \"inactive\" flag is NEVER SET, so \"registered on LessWrong\" is not evidence of anything. Here the EVENT LIST WITH DATES was the evidence, and it shows roughly two gatherings in 2026 — alive but slow. The undated \"Mar 15\" and \"Jan 18\" entries read as 2026 because LessWrong omits the year for the current year, but that is an inference about the site's date formatting and is labelled as such. The check was worth running and it did not change the picture: Atlantic Canada's only AI-safety entries remain the AI Atlantic Research Consortium and the Atlantic AI Institute (both adjacent) plus TrojAI in Saint John NB. Recorded so Atlantic Canada is not re-searched from scratch.", "editorial_scope": "excluded", "dedicated": false }, { "name": "Institute for AI Safety and Strategy (IAISS) — AI safety vocabulary with no verifiable output (checked 2026-08-31)", "slug": "institute-for-ai-safety-and-strategy-iaiss-ai-safety-vocabulary-with-no-verifiable-output-checked-2026-08-31", "website": "https://ai-rd.ca", "city": "", "province": "BC", "country": "Canada", "sector": "nonprofit", "kind": "", "focus": [], "status": "", "summary": { "en": "Uses core AI safety vocabulary with no verifiable output attached. Its own words: \"IAISS advances responsible AI by tackling safety in agentic and embodied systems... Key efforts target AI alignment, where systems behave as intended without unintended consequences\"; \"Policy considerations and recommendations in suggesting global coordination to dramatically reduce catastrophic AI risks\"; \"Collaboration with bodies like AI Safety Institutes.\" No named people, no dated activity, and no Canadian output of any kind was found. Its only shipped product is a three-part course on responsible generative-AI use in Saudi Arabian healthcare.", "fr": "" }, "key_people": [], "founded": "", "founded_year": "", "confidence": "negative finding", "last_verified": "2026-08-31", "sources": [ "https://ai-rd.ca" ], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "CHECKED 2026-08-31. Canadian status IS first-party — the ai-rd.ca footer reads verbatim: \"© 2026. All rights reserved. The Institute for AI Safety and Strategy (IAISS) is a registered non-profit organization in British Columbia, Canada.\" ⚠️ AIGS's \"Vancouver, Canada\" is NOT SUPPORTED and is not inherited: no city or address appears anywhere on either of the org's sites, so the location is BC province-wide and the city field is deliberately empty. WHY EXCLUDED DESPITE CORE VOCABULARY — nothing verifiable is attached to it: no named people, no founding date, no publications, no events, no dated activity of any kind, and the prose is generic throughout. The only contact is info@ai-rd.ca and the Thinkific /pages/about path 404s, so there is no about page. ITS ONLY SHIPPED OUTPUT is parts 1-3 of \"Responsible AI: Generative AI Use in the Kingdom's Healthcare System\", built around Saudi Arabia's PDPL, MoH/SDAIA/NHIC governance and Vision 2030 — real work, but neither Canadian nor AI safety in this map's sense. ITS \"TECHNICAL GOVERNANCE PLATFORM\" IS NOT DEPLOYED: technical-governance.netlify.app returns the title \"Technical International AI Governance Hackathon\" over an un-built React shell reading \"Enable JavaScript to run this app. Run: npm run dev, then open http://localhost:5173\" — the developer placeholder is live in production. AND ai-rd.ca IS NOT PRIMARILY IAISS: its is \"Sentinel Nexus: Secure AI Solutions for Government\", and the same page markets commercial products (Sentinel Nexus, Aurora Core, NexusInt, \"Rapidly Deployable, Secure, End-to-End AI Infrastructure\", \"OSINT Expertise\") to government buyers. A registered non-profit and a government AI-security vendor share one website with no stated separation. The URL AIGS carries is a Thinkific course-hosting subdomain, which is itself informative about the org's shape. RE-CHECK if named people or dated Canadian output ever appear.", "editorial_scope": "excluded", "dedicated": false }, { "name": "Queen's AI safety presence", "slug": "queen-s-ai-safety-presence", "website": "", "city": "Kingston", "province": "ON", "country": "Canada", "sector": "", "kind": "", "focus": [], "status": "", "summary": { "en": "No dedicated AI safety group confirmed. Queen's AI presence is Ingenuity Labs (AI, robotics and human-machine interaction, where \"safety\" means physical and human-factors safety) and QMIND, a large undergraduate applied-AI club. Neither addresses alignment or catastrophic risk.", "fr": "" }, "key_people": [], "founded": "", "founded_year": "", "confidence": "negative finding", "last_verified": "", "sources": [], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "CHECKED, nothing dedicated found. Closes the doc's open \"Queen's\" gap. Recorded so it isn't re-searched from scratch. Adjacent context: Queen's partnered with SFU in Mar 2026 on sovereign supercomputing capability, which is compute, not safety.", "editorial_scope": "excluded", "dedicated": false }, { "name": "Waterloo AI safety presence", "slug": "waterloo-ai-safety-presence", "website": "", "city": "Waterloo", "province": "ON", "country": "Canada", "sector": "", "kind": "", "focus": [], "status": "", "summary": { "en": "No dedicated student safety group confirmed. Waterloo appears via CIFAR's AI Safety Solution Networks (Maura Grossman) and the Cybersecurity & Privacy Institute's CIFAR special issue on \"AI Safety and Public Policy in Canada\".", "fr": "" }, "key_people": [ "Maura Grossman" ], "founded": "", "founded_year": "", "confidence": "negative finding", "last_verified": "", "sources": [], "scale": { "count": null, "unit": "", "bucket": null, "safety_count": null, "budget": "", "source": "", "checked": "" }, "notes": "CHECKED, nothing dedicated found. Recorded so the gap isn't re-searched from scratch.", "editorial_scope": "excluded", "dedicated": false } ], "counts": { "listed": 45, "checked_not_listed": 11, "verified": 23, "with_scale": 16 } }