# Frontier AI Governance course Canonical URL: Six Tuesday evenings to build the judgment to work on how frontier AI is governed. Based on the Frontier AI Governance course developed by BlueDot Impact; we follow their curriculum and meet in person in Montréal. [Apply by November 10 →](https://docs.google.com/forms/d/e/1FAIpQLSfnuvOnbU1PFzrsWprPTWnYmNoSd9lr9iNqgWl89ewHK2HpUQ/viewform) Not sure it’s for you? Apply anyway. - **When**: Tuesdays, 6 to 9 PM, November 17 to December 22, 2026 - **Where**: Ω Labs, 3813 rue Saint-Denis, Montréal - **Time commitment**: About 6 to 7 hours a week (4 to 5 hours of reading and exercises at home, 3 hours in person) - **Cohort**: 12 people - **Language**: English (readings and sessions); questions in French are welcome - **Facilitators**: Orpheus Lummis and Étienne Langlois - **Cost**: Free - **Apply by**: November 10 at 11:59 PM EST ## Who it’s for - People with a technical background (engineers, researchers, founders) who want to see how their skills could inform policy and institutions. - People from policy, law, security, economics, diplomacy or journalism who want informed judgment on frontier AI. - People weighing a next step into AI governance work, research or study, who want to test that interest with serious reading and discussion. No previous course or technical degree is needed. Some familiarity with what current AI systems can do helps. ## The six sessions Each week: do the unit’s core readings and exercise before Tuesday, then three hours of discussion at Ω Labs. ### Nov 17 [Unit 1: Where the frontier stands](https://bluedot.org/courses/ai-governance/1/1) How to read the evidence on frontier AI: system cards, independent evaluations, capability trends, and a 2026 case study, the OpenAI / Hugging Face incident. Exercise: a short briefing for a decision-maker. Core readings include: - Ajeya Cotra, “Takeoff speeds rule everything around me” (2026) - Allan Dafoe, “AI Governance: Overview and Theoretical Lenses” (2022) - Ji, Venkatram and Batalis, “AI Safety Evaluations: An Explainer” (2025) - METR, “Task-Completion Time Horizons of Frontier AI Models” (2026) - Helen Toner, “Taking Jaggedness Seriously” (2025) - Greenblatt, Cotra and Wijk on the Hugging Face incident (2026) - OpenAI, GPT-6 Astra System Card (2026) ### Nov 24 [Unit 2: Power, windows and dependencies](https://bluedot.org/courses/ai-governance/2/1) Who can act on frontier AI (companies, governments, international bodies), what each depends on, and where gaps and policy windows open. Exercise: map the actors. Core readings include: - International AI Safety Report 2026, extended summary for policymakers - METR, “Common Elements of Frontier AI Safety Policies” (2026) - Anton Leicht, “The Next Three Phases of AI Politics” (2026) - Matt Sheehan and Scott Singer, “How China Views AI Risks and What to do About Them” (2025) - Arnold and Llerena, “When Reporting an AI Security Incident Is Not Mandatory” (2026) ### Dec 1 [Unit 3: The proposal landscape](https://bluedot.org/courses/ai-governance/3/1) Competing strategies for governing frontier AI, from international agreements and compute governance to private governance and auditing, and the “Plan A” versus “Plan S” debate. Exercise: make the best case for a view you disagree with. Core readings include: - Christoph Winter and Charlie Bullock, “Radical Optionality” (2026) - Delaney, Kostylew, Guest and Wildeford, “Strategic Visions in AI Governance” (2026) - Narayanan, Ji, Venkatram and Luong, “AI Governance at the Frontier: Unpacking Foundational Assumptions” (2025) - Larsen et al., “AI 2040: Plan A” and “Plan S” (2026) - Scher et al., “An International Agreement to Prevent the Premature Creation of Artificial Superintelligence” (2025) ### Dec 8 [Unit 4: Governance under acceleration](https://bluedot.org/courses/ai-governance/4/1) What changes if AI speeds up AI research: race dynamics between companies and states, and the risk of power concentrating in few hands. Exercise: stress-test a proposal against fast progress. Core readings include: - MacAskill and Moorhouse, “Preparing for the Intelligence Explosion” (2025) - Hendrycks, Schmidt and Wang, “Superintelligence Strategy” (2025) - CSET, “When AI Builds AI” (2026) - Michael Nielsen, “ASI existential risk: reconsidering alignment as a goal” (2025) - Richard Ngo, “Distributed vs centralized agents” (2026) ### Dec 15 [Unit 5: Contested questions](https://bluedot.org/courses/ai-governance/5/1) The group picks one debate: how to handle open-weight models as they get more capable, or when, if ever, frontier development should be slowed and by whom. Write a position and defend it. Core readings include: - Kapoor, Bommasani et al., “On the Societal Impact of Open Foundation Models” (2024) - Ryan Greenblatt, “When is it important that open-weight models aren’t released?” (2025) - Joe Carlsmith, “On Restraining AI Development for the Sake of Safety” (2026) - Anthropic, Responsible Scaling Policy v3.4 (2026) - Nora Belrose, “AI Pause Will Likely Backfire” (2023) ### Dec 22 [Unit 6: Finding your leverage](https://bluedot.org/courses/ai-governance/6/1) Take stock of your skills and options, get peer feedback on a 30-day action and a six-month plan, and look at fellowships and roles (GovAI, Horizon, IAPS, Talos). Core readings include: - 80,000 Hours, “The US AI policy landscape: where to work to have the biggest impact” (2025) - Jordan Schneider, “Policy: An Early Career Guide” (2025) - Horizon Institute for Public Service, “Emerging Policy Careers: AI Policy” (2025) Each unit also lists optional readings. Each unit heading links to BlueDot’s unit page; a free BlueDot account unlocks the full reading guides. ## Selection Applications close November 10. We read every application and answer everyone by November 13; accepted people confirm their place. We look for people who will do the readings, come every week and act on what they learn. ## FAQ - **Do I need a previous course or a technical background?**: No. - **Is this about AI ethics or corporate compliance?**: No. The focus is frontier AI: the policy, coordination and institutional choices that shape how the most capable systems are built and deployed. - **Why in English?**: The readings are in English. You can ask questions and discuss in French. - **Can I miss a session?**: Tell us ahead of time. The course works best when everyone comes all six weeks. - **What should I bring?**: A laptop. Curriculum: the [Frontier AI Governance course by BlueDot Impact](https://bluedot.org/courses/ai-governance/1/1) (as updated September 2026), used under BlueDot’s guidance for independent cohorts. This is AI Safety Montréal’s own cohort, not an official BlueDot one. Readings belong to their authors.