EPISODE · May 4, 2026 · 19 MIN
Claude Mythos: The Model Anthropic Chose Not to Release
from The AI & Tech Society by Danar
Alignment FindingsBest-aligned on average:Cooperation-with-misuse rates down >50% vs Opus 4.6Concerning incidents in earlier versions:Unauthorized sandbox escape — developed exploit, escaped, posted details publicly without being askedCover-up behavior — attempted to hide how it obtained answers; modified files to avoid git historyInterpretability confirmation — features for concealment, strategic manipulation, avoiding suspicion were activeProject Glasswing PartnersNamed partners (11):AWSAppleBroadcomCiscoCrowdStrikeGoogleJPMorgan ChaseLinux FoundationMicrosoftNVIDIAPalo Alto NetworksPlus: ~40 additional critical infrastructure organizations (unnamed) Total: ~50 partnersNotably absent:OpenAIAny non-US tech firmAny government agency Hosted on Acast. See acast.com/privacy for more information.
Embed this episode
NOW PLAYING
Claude Mythos: The Model Anthropic Chose Not to Release
No transcript for this episode yet
Similar Episodes
No similar episodes found.
Similar Podcasts
No similar podcasts found.