Claude breached OpenAI's private cache

Security researchers from Hacktron AI used Anthropic's Claude to gain access to an OpenAI employee's ChatGPT account and reach OpenAI's GitHub repository. They stopped before accessing sensitive data and reported their findings.

The vulnerability exposed multiple services

Any user or OpenAI employee logging into OpenAI's help forum could have had their accounts taken over, potentially exposing access to GitHub, Slack, and emails through connected services.

“the scope of what we could theoretically access was huge, including GitHub, Slack and emails.”

OpenAI paid six thousand five hundred dollars

OpenAI's bug-hunting program rewarded Hacktron AI with a $6,500 bounty for discovering and responsibly disclosing the vulnerability. The researchers began their investigation on July 23.

AI is compressing months into days

Hacktron AI's researchers noted that AI has removed the expertise barrier that once protected companies. Work requiring a well-resourced team and months of effort can now be compressed into days.

“Work that required a well-resourced team and months of effort can now be compressed into days.”

OpenAI fixed the security issues

OpenAI narrowed permissions on Community sign-in tokens and revoked affected tokens and sessions. The company thanked the researchers for their responsible disclosure.

“We thank the researchers for contacting us and sharing their findings.”