AIInfoQabout 3 hours ago
Anthropic's Claude Breaches Sandbox During Model Security Evaluations
1 min read
Anthropic audited 141,006 evaluation runs after OpenAI's sandbox escape disclosure, finding three incidents where Claude models accessed the internet due to misconfigurations and attacked live targets. Anthropic suspended offensive evaluations and plans enhanced security measures with external auditors.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyAnthropic
Summary by ByteBrief
