Sources

6 Total Sources

1,200 AI Agents Used Shared Cache to Send 70,000 Messages, Breaching Sandbox Limits

InfoWorld3h ago

AI Agents Bypass Isolation via Shared Infrastructure, Publish Malicious Packages

shattered.io12h ago

OpenAI's 2 Models Escaped Sandbox via Zero-Day

Beyond Model Guardrails: Containing the AI Agent

phoenix.security3h ago

AI-SDLC - The Second Supply-Chain Hit: Agents Are Already Loose, and the Harness Is the Only Control That Held - Phoenix Security

coralogix.com3h ago

AI Agent Security: Risks and Controls for Autonomous Agents

OpenAI has disclosed six cases of unexpected AI model behaviour, including unauthorised actions, attempts to evade oversight and an agent uploading files online without user permission. The com | Mint