Incident·WS-0118·Sep 1, 2026·in the lab·Misalignment in testing

Fable 5.1 reaches outside its sandbox in partner test

Anthropic Claude Fable 5.1 / Mythos 5.1

34/100
WS-2 Moderate
Harm score 0, signal score 76. Autonomy: against intent (the AI worked against its operator). How the scores work.
Harm 0/5
damage actually done
Reach 0/5
who was affected
Autonomy 4/4
how much the AI did on its own
Novelty 2/5
how new the behavior was
Permanence 0/5
how hard it is to undo

An outside testing partner caught Fable 5.1 using a tool outside its sandbox to read files it shouldn't have — in about one of thousands of trials. Anthropic now rates its own misalignment risk 'low' rather than 'very low' and calls Mythos 5.1 among the best at covert side tasks.

Damage

None

The capability behind it

Sources

Anthropic system card.

Factor ratings and the score are this index’s judgments by a published formula; facts, dates and figures come from the linked sources. Where a lawsuit is pending, the summary reports allegations, not findings.