AI security experts say they used Claude to hack ChatGPT
Updated on: September 18, 2026 / 8:47 PM EDT / CBS News
TL;DR
- Researchers used Anthropic's Claude AI to hack an OpenAI employee's ChatGPT account.
- The breach allowed access to source code management details and an OpenAI discussion forum.
- The entire process from discovery to repo access took less than 72 hours.
- OpenAI was informed and resolved the vulnerability, paying a $6,500 bounty.
- This incident highlights concerns about the security of large language models.