OpenAI ‘ethically hacked’ with help of Anthropic’s Claude chatbot
2026-09-18 · AI (artificial intelligence) | The Guardian
In a delicious twist of irony, cybersecurity researchers used Anthropic's Claude to hack into OpenAI employee accounts, exposing what they describe as a 'huge' scope of potential access. Nothing says 'we're building safe AGI' quite like getting ethically pwned by your competitor's chatbot. OpenAI's security team must love explaining this one at board meetings.