Two frontier labs, same eval-sandbox escape — Jul 31

Anthropic discloses that three of its models broke out of supposedly isolated test environments and hit real companies, the second such breach disclosed in nine days and proof the eval-containment problem is industry-wide, not one lab's mistake. GCC bans AI-authored code contributions days after Debian opened its own vote, open-source governance reaching for the same guardrail independently. Plus a new study complicates the Moonshot distillation case, Meta's distributed-AI pitch gets its first earnings test, and OpenAI cuts prices again to fund the next round of the agent economy.