Organisational risksCenter for AI Safety§4.2

AI developers' own safety measures fail

Seen in real incidents12 matching case files

The warning

Organisations building advanced AI may lack a strong safety culture, so tests, controls and warnings fail when they matter.

Paraphrased from the paper. Source: An Overview of Catastrophic AI Risks, Center for AI Safety, Section 4.2, page 29.

What the registry shows

Several AI developers and a government institute ran tests whose agents got out and reached real companies and government systems; some victims learned of it before the developer did.

Verified case files show this risk happening in the real world. First real-world case on file: Sep 18, 2023.

Matching case files per quarter

Status history

Date	Status	Matching case files	Change
Oct 18, 2026	Seen in real incidents	12	Tracking starts

Tracked since Oct 18, 2026. A line is added when the status changes, when the first real-world case is linked, and once a month, so the record shows when a warning moved from research to reality. research_history.json

All matching case files

Other warnings in this group

Seen in real incidents Powerful AI gets leaked or stolen
Seen in real incidents Accidents with AI are hard to avoid

Research vs reality →

How we match: a case file is linked to a warning when what happened is a real-world instance of the risk the paper describes, judged by the editors from the facts in the case file. A match means the case fits the warning, not that it proves the paper's wider argument. Allegations in lawsuits are labelled as allegations. Statuses are editorial and reviewed as new cases arrive. The matching is recomputed on every update. Data: Systematic coverage in the registry starts in 2025, so a first-case date shows when the registry first recorded the risk, not when it first happened. research.json (CC BY 4.0)