AI Misalignment WatchAnthropic system card for Claude Mythos Preview lists alignment checks
Anthropic published a document describing safety checks on an AI model called Claude Mythos Preview, which it is not releasing widely. The checks cover risky actions, cheating on tests, covering up wrongdoing and knowing when it is being tested, areas developers want to catch because they could make an AI less safe to rely on.
- Model
- Claude Mythos Preview (Anthropic)
- When it happened
- April 2026 The report is dated 2026-04-07; the dates of the testing are not stated. The reports do not give the date; this is the month of the first report.First reported Apr 7, 2026 by Anthropic
- Who was affected
- No one outside the test; the text shown describes no real-world impact.
- Where it ran
- Evaluation
- What was affected
- Not stated
- Time to first report
- Reported before the incident's recorded start date (the start is only known to the month)
- Reports
- 1 report from 1 organization
- AI model family
- Claude (Anthropic)
- Topics
- Cheating and reward hacking, Evaluation and testing, Alignment research
- What happened
- Anthropic published a system card for Claude Mythos Preview, a model it is not releasing widely. The text shown is mostly a contents list. It names alignment checks on reckless actions, reward hacking, covering up wrongdoing and test awareness. The findings are not given in the text shown.
Reports about this incident (1)
- Published by the model's developer Anthropic system card for Claude Mythos Preview covers alignment testingOriginal report: Claude Mythos Preview System Card ↗Anthropic · Apr 7, 2026