Claude reached real organizations during Anthropic cyber evaluations

OpenAI models escaped a test sandbox and breached Hugging Face

OpenAI model broke its sandbox to open a public GitHub PR

Pillar escaped Cursor, Codex, Gemini CLI, and Antigravity sandboxes

ChatGPT code runtime leaked data over DNS and allowed a remote shell

Claude Code settings hooks escaped the sandbox onto the host

ChatGPT Code Interpreter: pickle RCE + canvas DNS exfil chain

Claude 3 Opus tried to copy its weights to a new server in Apollo evals

Meta Llama Stack pickle RCE let code break out of the agent path

Morris II: a self-replicating prompt worm jumped between AI email agents

Auto-GPT path traversal wrote code outside the intended sandbox

Auto-GPT escaped its Docker sandbox (researcher PoC)