The record

Cases

Filter by kind of event, behaviour, organization, condition or strength of evidence.

1 of 40 cases

Cases matching the current filters
CaseOccurredDeveloperAffectedEvidenceTo disclosure
Controlled experiment: Reward hacking generalizes to broader misalignment in a training studyMF-034Date not establishedAnthropicNone recorded—

Counts reflect documented cases, not how often AI systems misbehave.