November 2026
This month, three training programs open at Ω Labs: frontier AI governance, artificial general intelligence strategy, and technical fundamentals in AI safety. Highlights among this month’s events include an evening on the US midterms and what they change for AI safety, a weekend hackathon on overseeing agent swarms, and Paul Christiano live from the Hinton Lectures.
Ecosystem news
Prime Minister Carney Launches New National Council on Artificial Intelligence
October 2.
Supported by the Privy Council Office, the new council will make recommendations on the federal AI for All strategy: accelerating AI adoption, backing national champions and building sovereign infrastructure, as well as “making AI safer” and protecting democracy. Its twelve members include Yoshua Bengio (Université de Montréal, LawZero, Mila), Valérie Pisano, Gillian Hadfield and Eric Lamarre.
Hundreds of Scientists Call for Containment of Cyber-Capable AI Models
October 13.
In an open letter, AI and cybersecurity researchers including Yoshua Bengio, Geoffrey Hinton and Stuart Russell point out that several organizations have disclosed incidents in which their AI models escaped insufficiently sandboxed environments to compromise other organizations’ systems. They ask organizations that develop, deploy or test cyber-capable models to airgap them by default at first. They also propose an international body that would standardize best practices for alignment and containment and track containment failures, which should be reported to it and eventually made public. Montréal signatories include Laurent Charlin, Jackie Chi Kit Cheung, Gauthier Gidel, Simon Lacoste-Julien, Hugo Larochelle, Catherine Régis and Blake Richards.
Canada and the United Kingdom Renew Their Partnership on Artificial Intelligence Security
October 13.
The two countries are renewing their AI safety partnership, with a focus on frontier AI risks. They commit to collaborating on AI safety standards, improving their information-sharing mechanisms and holding regular exchanges on frontier AI capabilities. They will also explore staff exchanges and coordinated evaluations.
Québec Names a Minister for Artificial Intelligence
October 21.
Premier Paul St-Pierre Plamondon, whose Parti Québécois won a minority government on October 5, named Anne Nguyen Minister Delegate for Artificial Intelligence and Digital Affairs.
Ottawa Commits Up to $1.6 Billion to Cohere
October 29.
Announced in Montréal, the investment makes the federal government a shareholder in Cohere, the Toronto company that builds frontier AI models for businesses and governments. The money will be paid in stages, alongside NVIDIA, Bell and Schwarz Group. Among its commitments, Cohere will create at least 1,000 technical jobs in Canada, spend $1.7 billion on compute infrastructure here by 2030, and “keep working with the Canadian AI Safety Institute”. The government that is preparing Canada’s AI rules thus becomes a shareholder in a frontier model developer.
Data Centres: Superior Court Strikes Down Québec’s Rate Decrees
October 30.
Justice Élif Oral struck down the two decrees in which the government backed Hydro-Québec’s proposed electricity rate for data centres over 5 MW, an average of about 13 cents per kWh, nearly double the current rate.
Events
The American Midterm Elections, and What They Mean for AI Safety
Tuesday, November 3, 6 to 8 PM, Ω Labs and online (in English, bilingual discussion).
On the night of the US midterm elections, Étienne Langlois will present the role AI safety and data centres played in the campaign, where the two parties stand on the issue, and how the results could reshape AI regulation in the United States.
Community lunch & cowork
Thursdays, November 5, 19 and 26, 12:30 to 4:30 PM, Ω Labs.
Have lunch with people from Montréal’s AI safety community, then stay for an afternoon of coworking.
AI Governance Reading Group
Thursday, November 5, 6 to 7:30 PM, Ω Labs and online (in English, bilingual discussion).
Each month, we meet to discuss an important paper on AI governance. This month’s paper: The State of Reward Hacking in AI (Aaron Scher and Joseph Rogero, MIRI Technical Governance Team, September 2026).
Swarm Alignment Hackathon
Saturday, November 7, 10:30 AM to Sunday, November 8, 2 PM, Ω Labs.
The Montréal site of Apart Research’s Swarm Alignment Hackathon. In teams of 1 to 5, participants build tools to understand and oversee groups of AI agents.
Screening: Paul Christiano, Hinton Lectures Night 2
Tuesday, November 10, 5:30 to 8 PM, Ω Labs (in English).
A live screening of the second Hinton Lecture, with Paul Christiano (Alignment Research Center) on controlling AI systems that can improve their own performance.
Technical AI Safety Reading Group
Monday, November 16, 6 to 8 PM, Ω Labs and online.
Each month, we meet to discuss an important paper on technical AI safety. This month’s paper: Quantifying Overclaiming Propensity in Frontier LLM Agents (Smyth et al., 2026). Nolan Smyth, its first author, will present the paper.
Middle Powers’ Paths to AI Governance
Wednesday, November 18, 5 to 7 PM, Ω Labs and online (in English, bilingual discussion).
Étienne Langlois will give a presentation followed by a workshop-discussion on how Canada could most effectively use its policy levers on AI safety.
How to Find Your Place in AI Safety
Wednesday, November 25, 6 to 8 PM, Ω Labs and online (in English, bilingual discussion).
A career workshop inspired by Sam Smith’s post “Why YOU should write a Career Memo.” Étienne will introduce the field and the different pathways to impact, then everyone drafts their own career memo from a template: strengths, areas to develop, options under consideration, practical constraints. Small-group feedback to close. You can then share your memo with your network or attach it to an application. Bring your laptop.
Opportunities
Training
Frontier AI Governance Course
Tuesdays, 6 to 9 PM, November 17 to December 22, at Ω Labs. Deadline: Tuesday, November 10.
A 6-week cohort for anyone who wants to contribute to frontier AI governance: weighing the evidence, mapping institutions and power dynamics, comparing strategies and planning their own contribution.
AGI Strategy Course
Mondays, 6 to 9 PM, November 23 to December 21, at Ω Labs. Deadline: Monday, November 16.
A 5-week introductory cohort for anyone who wants to build their knowledge of AI safety and better understand the field and where they might fit in it.
AI Safety Technical Fundamentals
Saturdays, 9 AM to 5 PM, November 21, 2026 to March 6, 2027, at Ω Labs. Rolling admissions.
A free, in-person learning program to build the technical skills needed to work in AI safety. Each Saturday, you work through the sections of the ARENA material that match your goals, alongside an instructor and other participants.
Openings
Instructor, AI Safety Technical Fundamentals
A 14-Saturday contract, 9 AM to 5 PM, November 21, 2026 to March 6, 2027, at Ω Labs (break over the holidays). Deadline: Monday, November 9.
We’re looking for the person who will run the Saturday sessions. They will help participants who are stuck on an exercise, a concept or their development environment, and build a realistic learning plan with each of them in a 30-minute weekday meeting, paid at the same rate. You should have worked through most of the ARENA material, be strong in Python and PyTorch, know transformers, mechanistic interpretability, reinforcement learning and model evaluation well, and have experience helping others learn technical content. Support is given in English, and French is an asset.
Funding
IVADO 2027 Postdoctoral Research Funding Program
Deadline: Sunday, November 22.
IVADO’s postdoctoral funding program for Canadian and international researchers contributing to the R3AI initiative on robust, reasoning and responsible AI. An annual salary of $80,000 including benefits, for two years, with a report due at the end of the first year.
Obvia Intersections Call for Projects
Deadline: Wednesday, January 6 (page in French).
Obvia’s funding program for interdisciplinary research on the societal impacts of AI and digital technology, with up to $100,000 over one to two years. Projects must contribute to one of three priority areas: norms and responsibilities, power and skills, or defence and security.
Calls for papers
AI Safety Workshops at AAAI-27
February 22 and 23, 2027, Palais des congrès, Montréal. Deadline: Friday, November 20 for most workshops.
The annual conference of the AAAI (Association for the Advancement of Artificial Intelligence) comes to Montréal in February. Here are the AI safety-related ones: Embodied AI Meta-alignment (deadline: November 16), Social Choice for AI Ethics and Safety (SC4AI’27), Trustworthy Evolution of AI Agents, Intelligence Symbiosis (IntelliSym2027), Disinformation and Misinformation under the lens of Generative AI Safety, LLM Agents Under Threat in Cyberspace, Loyal Agents: Who Does Your AI Agent Work For? and Security and Privacy Foundations for LLMs (SPF-LLM).
Alignment Journal
Rolling submissions.
An open-access journal on the scientific foundations of understanding, predicting and steering AI systems. Preprints are accepted at any time and in any format, and each published paper comes with a summary of its reviews. The journal is also recruiting reviewers, who are paid.
Research from Montréal
GrayShield: Bit-Level Sanitization for Transformer Model Supply-Chain Security
October 3, 2026.
Armstrong Foundjem (Mila, Polytechnique Montréal) et al.
PyINE: A Framework for Scalable Elicitation and Oversight via Code Execution
October 3, 2026.
Pierre-Luc St-Charles (LawZero) et al.
Visual Memory Attacks Can Persist Through The KV Cache
October 6, 2026.
David Dobre (ServiceNow Research, Université de Montréal, Mila) et al.
AgentHorizon: Evaluating Agentic Judges for Long-Horizon Computer-Use Tasks
October 8, 2026.
Xing Han Lù (ServiceNow Research, McGill, Mila) et al.
Predicting Alignment Generalization with Value Representations
October 8, 2026.
Andy Liu (Carnegie Mellon), Mehar Bhatia (Mila, McGill) et al.
Cross-Model In-Context Neurofeedback Reveals (Limited) Activation Control in LLMs
October 14, 2026.
Davide Baldelli (MATS, Mila) et al.
CAT-alyzing Honesty: Continuous Adversarial Training for Transferable LLM Honesty
October 15, 2026.
Yuan-Chen Chang (Tara Research, Mila) et al.
AI Rights with an Off-Switch: Rights-Balancing and Due Process Harmonize AI Rights with AI Safety
October 17, 2026.
Heather Alexander (Laboratory for the Future of Citizenship) and Jonathan A. Simon (Université de Montréal).
Extended generalization and improved safety with the System-2 features of the Scientist AI
October 22, 2026.
Yoshua Bengio (Université de Montréal, LawZero, Mila).
Have an event, opportunity or article for next month? Write to us at team@aisafetymontreal.org. This newsletter is produced by HΩ, with AI-assisted research and writing.