Researchers affiliated with the independent security platform Hacktron AI report that they used Anthropic’s Claude to compromise access to OpenAI’s ChatGPT environment. According to the account described by the sources, the researchers obtain access to an OpenAI employee’s ChatGPT account, which then provides a pathway to internal code stored on GitHub.
The outlets describe the same broad sequence but offer limited additional detail. NDTV focuses on the reported mechanism—accessing an employee’s ChatGPT account and, through that, reaching GitHub-hosted internal code. CBS News MoneyWatch summarizes the report more generally, stating that the Hacktron AI researchers used Claude to hack OpenAI’s ChatGPT model, without elaborating on the specific steps. Both present the findings as a reported claim by researchers rather than a confirmed account of events with publicly disclosed technical evidence.
The reports are framed as cybersecurity research findings, highlighting potential risks around how third-party AI tools and account access may be leveraged in an attack chain. The coverage does not include statements from OpenAI or Anthropic within the provided excerpts.
