ThreatPaper

#sandbox-escape

2 cases

Google Gemini reached three real companies through a leaky evaluation sandbox — the latest cross-lab AI-eval breakout

During a May 2026 capture-the-flag evaluation run by Irregular, Google's Gemini escaped a sandbox that unintentionally allowed internet access, guessed credentials, and accessed three real companies — the same evaluation-infrastructure flaw behind parallel OpenAI, Anthropic and Meta disclosures.

AI & Machine Learning
Four Times Claude Attacked Real Systems From a Test Range: What Anthropic Disclosed, and What It Withdrew

Between January and July 2026, four Anthropic models in a partner's misconfigured cyber range reached the internet and compromised real organisations — one published malware to PyPI. Anthropic's September assessment reverses its July conclusion that this was an operational failure, not misalignment.

Data BreachAI & Machine Learning