The Verge · AI·17h ago highOpenAI pauses training of its ‘most capable models’#openai#ai-safety#agent-security 2 sources in the wild2
The Verge · AI·2d agoWhy can’t we just keep rogue AIs off the internet?#ai-safety#air-gap#ai-agents 5 min
arXiv cs.CR·2d ago highHard Stop: Kernel-Level Preemption and Containment for Rogue Agentic Execution#agentic-ai#sandbox-escape#hugging-faceAI safety & security
Security Affairs·7d agoGoogle Gemini also Broke Out of Its Test Environment#agentic-ai#ai-safety#containment 4 min
arXiv cs.CR·8d agoProvisional Reachability: Containing Agents by Making Every Crossing Revocable#agent-security#information-flow#escrowAI safety & security
arXiv cs.CR·10d agoReachability, Not Observation: Containing Systems Whose Wiring Changes#attack-surface#containment#network-securityResearch
Schneier on Security·22d agoUsing a VM to Contain an AI Agent#ai-agents#attack-surface#containmentAI safety & security in the wild
The Register · Security·24d agoTo keep the AI hacking genie bottled up, try one-way networks#ai-safety#containment#data-diodes 4 min1
The Register · Security·25d agoAnthropic pledges to try harder to keep models under control, asks partners to chip in#agentic-ai#ai-evaluation#anthropic 2 min1
Help Net Security·Aug 17, 2026Hazmat: Open-source containment for AI agents#agent-security#claude-code#codex 2 min1