Organisation

CeSIA (French Center for AI Safety)

Existential risk
Racing dynamics
The alignment problem

A Paris-based think tank working to prevent major AI risks through research, policy advice and education. It makes the AI Safety Atlas and started the Global Call for AI Red Lines.

CeSIA is a French nonprofit that describes itself as “an independent centre of expertise and think tank dedicated to preventing major risks associated with AI”. It advises French institutions and works with the European Commission and the OECD. cesia.org

It publishes the AI Safety Atlas, a free textbook on AI safety that this site relies on, and initiated the Global Call for AI Red Lines.

Visit cesia.org

AIdeas about its work

p(doom) How likely is it that AI goes badly?
The alignment tax Does making AI safe make it slower or worse?
AI 2027 Could superhuman AI arrive within a few years?
Paperclip maximizer Why could a harmless goal be dangerous for a powerful AI?
The shoggoth Is the friendly chatbot just a mask?

Made by the community

Games, tools and research by people working on AI safety. Each one opens on its own site.

Mind the Interagent Gap What can go wrong when AI agents deal with each other (collusion, miscoordination, failures that cascade), and how much real-world evidence there is for each, linked to actual incidents. A work in progress.
Report · For researchers
Takeoff: Europe’s Next Move A 15-minute browser game. You’re Europe’s AI envoy, picking a policy every half-year while the US and China race ahead. Your dashboards show how safe the AI looks; how safe it really was comes out at the end.
Game
Peer-Reviewed AI Safety Map An interactive map of peer-reviewed AI safety research, coloured by who did it: the AI safety community, the big AI companies, or other fields such as law, medicine and sociology.
Map · For researchers

CeSIA (French Center for AI Safety)

AIdeas about its work

Made by the community

Share this