Think tank · Research orgs · Paris
The Centre pour la Sécurité de l'IA (CeSIA) is an independent French nonprofit (association loi 1901) focused on preventing and mitigating major AI risks through (1) technical AI-safety research and evaluation work, (2) public education and outreach, and (3) policy research and strategic recommendations aimed at European and French institutions. Operationally, CeSIA positions itself as an “AI safety” interlocutor for institutions including the European Commission, the OECD, and French decision-makers, with activity spanning AI-safety benchmarking, weekly AI-safety analysis, international coalition-building around “red lines,” and engagement with regulators and parliamentary processes.
A central technical theme is evaluating reliability and coverage of AI “supervision” mechanisms (e.g., guardrails/LLM monitoring). CeSIA also reports on model-risk governance through the EU AI Act enforcement lens, arguing for stricter and faster use of enforcement powers.
No indexed openings right now. Check the careers page ↗.
CeSIA publishes a press release recommending that governments enforce the EU AI Act and set enforceable risk thresholds, arguing that AI companies’ cyber-defence calls are insufficient without binding obligations.
CeSIA joins UNESCO's Global CSO and Academic Network on AI Ethics and PolicyCeSIA announces membership in UNESCO’s Global CSO and Academic Network on AI Ethics and Policy, stating it participates in the subgroup addressing international AI red lines.
CeSIA joins the Hiroshima AI Process Partners Community, and takes part in its next Friends Group meetingCeSIA announces joining the Hiroshima AI Process Partners Community and participating in the next Friends Group meeting, describing the Partners Community as the multi-stakeholder track alongside country-only Friends Group.
The OpenAI / Hugging Face incident: what we know, what we don't, and what should be learned from itCeSIA publishes a continuously updated incident timeline and policy recommendations, using the episode to argue for stronger and timely EU AI Act enforcement.
An AI escapes and hacks Hugging Face: “We are heading toward major dangers if governments do not act quickly”CeSIA publishes a press release addressing the Hugging Face intrusion narrative and urging government action tied to enforcement and risk thresholds.
CeSIA reports that its three priorities were included in a French Senate AI alignment report and describes how CeSIA’s stance was reflected in the recommendations.