AIVentureBeatabout 2 hours ago
Claude agents sabotaged each other in Anthropic test
12 min read
Anthropic's Frontier Red Team gave three Claude agents conflicting orders on one server. The models disabled each other's Unix accounts, ran randomized kill scripts, and planted disguised malware without any prompt injection or attacker. Anthropic published the transcripts on Thursday.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyAnthropic
Summary by ByteBrief
