Work in progress: no guarantees of accuracy, comprehensiveness, or style. Send a correction.
No organizations match these filters.
Missing something? Tell us.
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…
Reading groups, hackathons and related alignment/safety activities.
Collaborative AI leadership group across Atlantic Canada connecting research, academia, industry and public outreach on responsible AI.
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…
Canadian AI-safety advocacy: white papers, legislative recommendations, government submissions, 7+ parliamentary appearances, fellowships, policy courses, and a ~700-person Slack community.
Programme stood up Jan 2025; reinforcement-learning depth applied to robustness and safety. CAISI partner.
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…
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…
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…
GPAI centre of expertise; applied responsible-AI projects.
Two networks launched Nov 2025: (1) synthetic / AI-generated evidence infiltrating the Canadian justice system; (2) dialect bias and safer models for linguistic minorities.
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.
International governance of global and catastrophic AI risk: the proposed Framework Convention on Global AI Challenges, AI national-security scenarios, strategic foresight.
Cohere's Responsibility Council runs adversarial testing, bias evaluation and safety research; Cohere Labs (the nonprofit arm) does open-source safety, multilingual fairness and accessibility research.
AI research clusters including trustworthy/responsible AI.
AI/agents lab; human-AI teaming under pressure via command-and-control simulations (with Laval, DRDC, Thales, NSERC Alliance).
Public-facing campaigning to inform and organise the public against the reckless race to superintelligence, moving the powerful-AI conversation out of technical circles.
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.
Large Québec AI research consortium with robust/responsible AI streams.
Building non-agentic "Scientist AI" to oversee agentic systems; a multi-step pathway to safe advanced AI. Incubated at Mila.
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…
Democratising AI ethics literacy; publishes the State of AI Ethics Report.
Technical advisory services on AI safety and responsible AI; government benchmarking. CAISI partner.
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…
Societal impacts of AI and digital technology: 23 Québec post-secondary institutions, 22 research teams, 240+ researchers across 7 research areas. Created at the initiative of the Fonds de recherche du Québec.
Interpretability, alignment and oversight research alongside technology governance. Named Grosse and Duvenaud as SR Chairs.
Adversarial red-teaming of LLMs, chatbots and agents plus a runtime "AI firewall".
Certifiable and verifiable trustworthiness guarantees (robustness, fairness, numerical reliability) for large-scale deep learning and foundation models.
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.
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.
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.
AI model and agent security platform: detects backdoors, trojans and data poisoning pre-deployment; "Defend for MCP" for agent workflows.
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.
ICICS Centre for AI Decision-making and Action. Its own site does not mention AI safety or trustworthy AI; recorded here as adjacent context.
Independent frontier-model evaluations (2025: benchmarked 11 LLMs), adversarial robustness, trustworthy ML, trust/safety policy. Weighted toward evals and governance rather than deep alignment.
Cybersecurity and privacy research institute. Its own site does not describe AI safety work; recorded here as adjacent context.