AI Safety Events · Text version

Week 46, 2026

This issue as simple text with links, ready to paste anywhere.

View the email

AI Safety Events · Week 46, 2026

This is a weekly newsletter that lists newly announced events addressing existential risk from AI. Visit AISafety.com/events for the full directory of upcoming events.

New events

• San Francisco Alignment Workshop 2027
Workshop · San Francisco, USA · 24 – 25 Apr 2027
By FAR.AI · Pay to attend · Apply by 24 Apr 2027
Gathers ML researchers and practitioners from industry, academia, and government for talks, discussions, and 1-1s. Topics include model evaluations, interpretability, robustness, scalable oversight, and AI governance.

• Technical Innovations for AI Policy (TIAP) Washington D.C. Conference 2027
Conference · Washington D.C., USA · 15 – 16 Apr 2027
By FAR.AI · Free · Apply by 15 Apr 2027
Brings together policymakers and technical experts from government, industry, think tanks, academia, and civil society. Covers frontier AI capabilities and where research can inform live policy debates.

• How can AI kill us all? Eight existential risk scenarios
Talk · San Francisco, USA · 19 Nov 2026
By Existential Risk Observatory (ERO) · Free · Register by 19 Nov 2026
Geoffrey Hinton gives a keynote and Q&A, and Buck Shlegeris, Daniel Kokotajlo, and Malo Bourgon join a panel. Hugo Save introduces the scenarios, including three takeover scenarios and a handover scenario.

• AI Safety Forum 2027
Conference · Online & Sydney, Australia · 6 – 7 Jul 2027
By Gradient Institute · Pay to attend (assistance available)
Brings together people from research, government, industry, and civil society to shape Australia's role in AI safety and governance. Includes presentations, workshops, and structured discussions.

• YC Debates: Should Governments Intervene to Slow Down AI Development?
Talk · San Francisco, USA · 16 Nov 2026
By Y Combinator · Free · Apply by 16 Nov 2026
Jacob Coxon, who left pretraining research at OpenAI and Anthropic to warn about the race to superintelligence, argues in favor. Economist Robin Hanson, who wrote that regulations to pause AI look risky, argues against.

• Provably Secure: Math Against Superintelligent Hackers
Talk · Cambridge, USA · 17 Nov 2026
By MIT Schwarzman College of Computing · Free · Register by 17 Nov 2026
Panel on whether formal verification can guarantee AI agent containment and find all security bugs of certain kinds. Covers AI as attacker, AI as a system to confine, and trusting specifications for agent-written code.

• Deep dive: Recursive self-improvement, and what it may mean for humanity
Talk · Berlin, Germany · 17 Nov 2026
By AI Safety Berlin, AUREA · Free · Apply by 17 Nov 2026
Julian Schulz of Principles of Intelligence and Meridian Cambridge shares insights from his work and discusses the possibility that recursive self-improvement arrives soon. Followed by Q&A, discussion, and networking.

• Middle Powers' Paths to AI Governance
Talk · Online & Montréal, Canada · 18 Nov 2026
By AI Safety Montréal · Free · Register by 18 Nov 2026
Étienne Langlois presents how Canada could best use its policy levers on AI safety, followed by a workshop and bilingual discussion in French and English. The format adapts to the level of expertise in the room.

• 2026 AI Policy Competition
Competition · Online & Charlottesville, USA · 15 – 22 Nov 2026
By Virginia AI Security Initiative (VAISI) · Free (cash prize available) · Register by 15 Nov 2026
Undergraduate and graduate students, solo or in teams of up to three, write a two-page AI policy brief with a recommendation, evidence, and projected implementation. Top six teams present to a panel of judges.

• AI Verification x Crypto
Meetup · London, UK · 13 Nov 2026
By Encode · Free · Apply by 13 Nov 2026
Covers what AI verification is and how engineers and researchers in zero-knowledge proofs, secure enclaves, security, or hardware can help build its tools. Includes an intro talk, a panel, and networking.

• Alignment from Token Zero
Talk · Zurich, Switzerland · 17 Nov 2026
By Zurich AI Safety · Free · Apply by 17 Nov 2026
V. Moskvoretskii presents Synthetic Persona Pretraining, which builds an assistant persona into a model from the start of pretraining by annotating documents with moral reflections. Followed by open discussion.

• Anthropic AI Safety Q&A: UT Austin x Rice x Texas A&M
Talk · Austin, USA · 17 Nov 2026
By Supervised Program for Alignment Research (SPAR) · Free · Register by 17 Nov 2026
Fabien Roger, a researcher at Anthropic, takes questions on his work and how the company approaches AI safety. One hour of Q&A, then optional 15–30 minute office hours with him. Not recorded and closed to press.

• Agent Attribution Hackathon
Hackathon · Online · 11 – 13 Dec 2026
By Apart Research · Free (cash prize available) · Register by 11 Dec 2026
Teams build tools that tell whether a human, or nobody, is steering an AI agent behind an incident. Tracks cover attribution classifiers, agent setups that evade attribution, and finding rogue agents in the wild.

• Quantifying Overclaiming Propensity in Frontier LLM Agents
Other · Online & Montréal, Canada · 16 Nov 2026
By AI Safety Montréal · Free · Register by 16 Nov 2026
Reading group on a paper measuring how often frontier coding agents claim work their own transcripts show they never did, presented by first author Nolan Smyth. Bilingual discussion in French and English.

• Building an organisation and how Africans can contribute to AI safety
Talk · Online · 17 Nov 2026
By Effective Altruism South Africa · Free · Register by 17 Nov 2026
Cecil Abungu, coordinator of the ILINA Program, speaks on founding and leading a high-impact organization in Africa and how Africans can contribute to AI safety. For aspiring founders and early-stage nonprofit leaders.

• Introduction to AI Safety – Talk with Safe AI Germany (Kaiserslautern)
Talk · Kaiserslautern, Germany · 25 Nov 2026
By AI Safety Kaiserslautern, Safe AI Germany (SAIGE) · Free · Register by 25 Nov 2026
Johannes Koch covers how fast AI is progressing and the risks posed by advanced AI. Includes the main research directions, the current state of the field, and how people with different skill sets can contribute.

• Science Fiction as Scenario Planning: The Gentle Romance by Richard Ngo
Other · Online · 20 Nov 2026
By Institute for the AI Century · Free · Apply by 20 Nov 2026
Book club discussion comparing stories from the collection, including one in which two people choose to share a mind. Asks how choices about autonomy and control might scale into lasting losses of human agency.

• AI Safety Reading Group
Other · Saarbrücken, Germany · 17 Nov 2026
By AI Safety Saarland · Free
Meets weekly to read together for an hour, then discuss for an hour. Topics include alignment, interpretability, evals and red teaming, AI control, deception and scheming, and governance. Members vote on readings.

• PauseAI Israel Call
Meetup · Online · 19 Nov 2026
By PauseAI · Free · Register by 19 Nov 2026
Casual session for people concerned about catastrophic risk from AI, its rapid development, and lack of oversight. Explores whether the country could have its own PauseAI group, with strategy and next steps for action.