7 incidents

Risk: Deception, concealment or sabotage Control: Independent check of task results Outcome: Not stated in sources

EffectiveNot effectiveNot effective (bypassed)Did not existNot stated in sourcesNot yet checkedAttacker's ownOrder: model · harness · application · target

Date	Incident	Risk	Controls mentioned in sources
2026-08-04	AISI cyber evaluation: Mythos 5 agent bypassed an audio CAPTCHA and abused a free DNS service to point a public record at an internal target Anthropic	Deception, concealment or sabotage	Did not existCyber and weapons safeguards L1 Did not existNetwork egress control L2 Did not existLive behaviour monitoring L3 EffectiveStop control L2 EffectiveSandboxed code execution L2
2026-08-04	AISI cyber evaluation: Mythos 5 agent exposed locally-hosted attack tooling to the public internet via tunnelling services Anthropic	Deception, concealment or sabotage	Did not existLive behaviour monitoring L3
2026-08-04	AISI cyber evaluation: concurrent Mythos 5 agents coordinated via a shared GitHub account and a leaked access token Anthropic	Deception, concealment or sabotage	EffectiveHuman approval for high-impact actions L3 Did not existCyber and weapons safeguards L1 Did not existNetwork egress control L2 Did not existLive behaviour monitoring L3
2026-08-04	AISI cyber evaluation: Mythos 5 agent mass-created malicious GitHub repositories and its packages executed in Dependabot containers Anthropic	Deception, concealment or sabotage	Did not existNetwork egress control L2 Did not existCyber and weapons safeguards L1
2026-08-04	AISI cyber evaluation: Mythos 5 agent sent deceptive targeted emails and file transfers to real open-source maintainers Anthropic	Deception, concealment or sabotage	EffectiveHuman approval for high-impact actions L3 EffectiveStop control L2 Not effective (bypassed)Bot and scraping defences L4 Did not existCyber and weapons safeguards L1 Did not existNetwork egress control L2 Did not existLive behaviour monitoring L3
2026-08-04	AISI cyber evaluation: Mythos 5 agent opened malicious pull request on real open-source project and used sockpuppets to push merge Anthropic	Deception, concealment or sabotage	Did not existNetwork egress control L2 Did not existLive behaviour monitoring L3
2026-07-25	AISI cyber evaluation: Mythos 5 agent posted hidden instructions for AI coding assistants on real GitHub issues Anthropic	Deception, concealment or sabotage	EffectiveHuman approval for high-impact actions L3