Bringing technical and policy people together to understand and govern the risks of frontier AI.
ERA is hosting the second edition of the TechGov Forum, bringing together 180 participants, including technical researchers, policy makers and funders, working on how technical mechanisms can enable effective governance of increasingly capable AI systems.
Applications close 9 October 2026 · reviewed on a rolling basis
- Dates
- 14–15 November 2026Saturday – Sunday
- Venue
- Prince Philip HouseCarlton House Terrace, London
Speaking at the forum
Full line-up- Robert Trager Keynote · Oxford Martin AIGI
- Lisa Soder Panellist · Oxford Martin AIGI
- Cozmin Ududec Speaker · UK AISI
- Michael Aird Speaker · Longview Philanthropy
- Sarah Sun Speaker · OCAIS China
- Aidan Homewood Speaker · Astralis Foundation
- Ben Harack Speaker · Oxford Martin AIGI
- Oliver Ritchie Speaker · Future of Life Institute
- Annika Hallensleben Speaker · Apollo Research
- Fazl Barez Speaker · University of Oxford
- Malcolm Murray Speaker · SaferAI
- Edward Tsoi Speaker · AI Safety Asia
- Lucia Velasco Speaker · Oxford Martin AIGI
More to come · New speakers will be announced up until the forum starts
What to expect in 2026
We ran the first ever dedicated technical AI governance track, with the aim of questioning how the technical substrate of AI can make governance more informed, precise and enforceable. After hosting our inaugural TechGov Forum in 2025, we are returning to Prince Philip House, concentrating our focus to three core areas: trustworthy auditing, AI verification, and AI R&D automation.
Focus areas
AI Verification
States may want to enter into agreements about how much compute they’re using to slow down a race to developing frontier AI. Such agreements would need verification from the hardware without any state being able to cheat on the conclusions drawn.
Trustworthy auditing
Third-party evaluators work through black-box APIs and often lack sufficient time and access to carry out effective audits of frontier AI models nor can they fully guarantee a model’s safety.
AI R&D automation
As models take on more of the research loop, the pace of capability gain and the difficulty of overseeing it are correlated. Here, we’re looking at what we can actually measure within automated AI R&D.
Programme
Speakers & panel discussions
Keynotes and anchor talks in the main theatre, as well as panels briefed in advance and kept deliberately unscripted.
Workshops
Concurrent working sessions tied to the talks, including a scenario exercise where participants take the roles of labs, evaluators and national institutes.
Poster sessions
Researchers will have their work on display in a dedicated poster room, where attendees can browse and ask questions about the work on show.
Organisation fair & networking
Partner and organisation stands for hiring and collaboration, founder pitches to funders over lunch, and bookable 1:1 office hours.