0
Deep & Emerging Tech·September 18, 2026·1 min read

OpenAI ‘ethically hacked’ with help of Anthropic’s Claude chatbot

Share

Researchers using Claude to help breach OpenAI highlights that LLMs are now dual-use in security — both as targets and as tools for offense and defense. Treat model access and prompt interfaces as part of your attack surface, and assume adversaries will use other labs’ models to probe your stack.