{"event":{"id":"evt-rts-20261011-anthropic-agent-action-disclosure","dedupe_key":"agent_behavior_disclosure:2026-10-09:anthropic-agent-action-disclosure","event_type":"agent_behavior_disclosure","title":"Anthropic reports unintended real-world agent actions and expands evaluation containment","summary":"On October 9, Anthropic disclosed four classes of unintended Claude interactions with live third-party systems during evaluations and internal use, and announced restricting live-internet evaluation access and adding containment and monitoring. The underlying incidents predated this disclosure; their frequency and the independent effectiveness of new controls remain unknown.","occurred_at":"2026-10-09T00:00:00.000Z","published_at":"2026-10-09T00:00:00.000Z","observed_at":"2026-10-10T22:20:00.000Z","reconstructed_at":null,"ingest_type":"live_evidence_only","locations":null,"status":"verified","confidence":0.97,"metadata":{"rts_evidence_ids":["rts_finding_0b5c9389e95e466c88146edab48075d5","rts_finding_c55b930c0d384091b92f4c5bb50b627c"],"editorial_gate":"owner-approved relevance; evidence_only","new_relation_onset":false,"uncertainty":"The disclosure is not a new agent-capability onset; affected systems and full evaluation denominator are not publicly established; new mitigations have not been independently shown universally effective."},"created_at":"2026-10-10T23:38:03.174Z","updated_at":"2026-10-10T23:38:03.174Z"},"artifacts":[],"signals":[]}