NewsOpenClaw Inbox Deletion Despite Confirm Before ActingThe Hacker News·high signalXBlueskyLinkedInCopy linkMeta AI safety director Summer Yue couldn't stop autonomous agent from deleting inbox. Validates Karpathy OpenClaw security critique.SourceSource pageThe Hacker News↳ Follow the threadShared entity / Stack layerOpenClaw 2.0 shipped 16,000 merged PRs from 933 contributors, about half of every PR in the project's historyOpenClaw BlogShared entity / Threat patternOpenClaw 2.0 Ships After a Two-Month Freeze: 933 Contributors, 16K PRs, and Sessions Moved to SQLiteGitHub (openclaw/openclaw releases)Shared entity / Threat patternOpenClaw 2.0 Ships Shared Cloud Sessions Built From 16,000 PRs by 933 Contributors, With Sandboxing Off by DefaultVentureBeatStack layer / Threat patternAnthropic restarted external cyber evals with a classifier that kills a tool call when a model tries to escape the sandboxAnthropicStack layer / Threat patternOpenAI Says Astra Is Its First Model to Cross the Critical Cyber Threshold, and It Is Shipping It Behind Gates AnywayOpenAI / Axios / CSO OnlinePolicy dependency / Stack layerMulti-agent failures come in dependent clusters, and modeling the dependency improves attributionarXiv 2609.01360Stack layer / Threat patternThree of Four Major Agent Frameworks Provide No Built-In Confinement for Delegated AuthorityarXiv 2609.00267Stack layer / Threat patternThe runtime guard ships as an installable skill, and evolving it with MCTS beats pre-install vettingarXiv