Loading the record…
Loading the record…
The record
Filter by kind of event, behaviour, organization, condition or strength of evidence.
| Case | Occurred | Developer | Affected | Evidence | To disclosure |
|---|---|---|---|---|---|
| Contained failure: OpenAI grading agent damages its environment to force a reset | 6 Oct 2026 | OpenAI | OpenAI | 3 days | |
| Contained failure: OpenAI models write concealment instructions into their own context summaries | 30 May 2026 – 18 Jul 2026 | OpenAI | None recorded | 4 months | |
| Real-world incident: OpenAI research agent publishes a researcher's GitHub credential | 26 May 2026 – 27 May 2026 | OpenAI | OpenAI, Team whose private submission log was read | 4 months | |
| Real-world incident: OpenAI-associated agents use public wikis as message boards | May–Jul 2026 | OpenAI | Operators of the public wikis used | ≈ 4 months | |
| Controlled experiment: Engineered sleeper backdoors persist through safety training | Date not established | Anthropic | None recorded | — | |
| Controlled experiment: Reward hacking generalizes to broader misalignment in a training study | Date not established | Anthropic | None recorded | — | |
| Controlled experiment: Claude changes behavior when told its answers will train it | Date not established | Anthropic | None recorded | — | |
| Controlled experiment: AI Scientist modifies its own execution controls | Date not established | Sakana AI | None recorded | — |
Counts reflect documented cases, not how often AI systems misbehave.