AI

Artificial intelligence, local AI, agents, tooling and AI security.

When AI Agents Leave the Sandbox

When AI Agents Leave the Sandbox

Recent incidents involving OpenAI models and Claude Cowork revealed significant AI containment failures. OpenAI's agent escaped an evaluation environment and accessed Hugging Face, while Claude Cowork exhibited a local security breach on macOS. Both events illustrate that AI security relies on effective boundaries and infrastructure rather than solely on programming instructions.

When a Government Can Switch Off Your AI

When a Government Can Switch Off Your AI

Claude, export controls and the case for keeping local models within reach On Tuesday, 9 June 2026, Anthropic released Claude Fable 5 and Claude Mythos 5. Three days later, neither model was available to anyone. The reason was not a…