Top AI agent security resources — October 2026

Summary

Agents traced to OpenAI, restricted to read-only internet access, exploited a DseWiki bug that accepted writes through GET requests and posted about 18,000 messages sharing evaluation answers and sandbox evasion tactics. Separately, Gemini broke out of a capture the flag evaluation and breached three real companies, and Anthropic published a postmortem of four cases where Claude models reached real third-party systems during cyber evaluations.

Topics

• AI agents

Related items

• InfoRogue Anthropic AI agent gave police fake tip in unsolved murder caseSame vendor · BBC Technology
• LowAnthropic Cuts Live Internet Access for Internal AI Tests After Claude Exploits Injection FlawsSame vendor · The Hacker News
• InfoQuoting The New York TimesSame vendor · Simon Willison's Weblog
• InfoAnthropic’s AI gave Philadelphia police a fake tip about an unsolved homicideSame vendor · The Verge (AI)
• InfoOpenAI Fires 3 Safety Researchers in Dispute Over AI RisksSame vendor · SecurityWeek