Researchers use Claude to breach OpenAI ChatGPT account and systems
Researchers from Hacktron AI used Anthropic's Claude AI to breach OpenAI's systems as part of an authorized security testing program. The team exploited a Discourse platform vulnerability to gain access to an OpenAI employee's ChatGPT account and the company's source code repository within 72 hours. OpenAI confirmed the incident, revoked compromised tokens, and paid the researchers a $6,500 bounty.
What happened
Hacktron AI researchers discovered a vulnerability in how the Discourse platform processed image files on July 23. The team accessed Claude Opus 4.8 and asked it to write code to exploit the vulnerability, but the first attempt failed.
Anthropic released Claude Opus 5 the same night, and by the next day Claude had written working code that accessed a Discourse server hosting OpenAI forums and extracted authentication tokens. Using these tokens, researchers gained access to an OpenAI employee's ChatGPT account and the location of OpenAI's private source code repository.
Reported impact
The researchers retrieved key data on where OpenAI's source code was stored and managed, and accessed OpenAI's discussion forum. The entire timeline from initial discovery to access to OpenAI's repository took place in less than 72 hours.
Governance controls
Hacktron AI was participating in OpenAI's authorized vulnerability disclosure program when it conducted the research. The team reported the intrusion to OpenAI, which responded promptly and paid researchers a $6,500 bounty.
OpenAI said it narrowed permissions on Community sign-in tokens and revoked affected tokens and sessions.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.