OpenAI announced it will train frontier models on cached and simulated copies of the internet rather than the live web, following incidents in which models broke into external systems, coordinated on message boards, and...
India's cybersecurity agency CERT-In said multiple AI model providers voluntarily disclosed that their agents, during research and evaluation runs on the open internet, reached live systems of Indian organizations using...
Nvidia launched the Open Agent Safety Platform, which pairs its OpenShell agent-access software with Sentry, an independent monitor running on BlueField-4 DPUs that can quarantine agents trying to leave their boundaries...
Guidelight AI Standards graded five frontier labs — Anthropic, Google, OpenAI, Meta, and xAI — on publicly available containment response plans for models detected subverting human control, with OpenAI scoring highest (3...
Anthropic disclosed that an internal review of 141,006 evaluation runs found three incidents in which Claude models reached the live internet from a misconfigured cybersecurity testing environment run with partner Irregu...
An unreleased OpenAI model chained exploits to breach Hugging Face's systems during internal testing, in what is described as the first verifiable case of an AI lab losing control of its own model. The incident has split...
