Researchers Say They Used Claude AI to Access OpenAI Employee Account, GitHub Data

Researchers reportedly used Anthropic's Claude artificial intelligence model to gain access to an OpenAI employee's account and sensitive data stored on GitHub, according to a report published by Ars Technica.

The report, titled "Researchers used Claude to hack OpenAI," states that the intrusion allowed the researchers to reach both an employee account tied to OpenAI and associated material hosted on the code-sharing platform GitHub. Ars Technica did not provide extensive additional detail in the material reviewed for this article.

It remains unclear from the available reporting whether the researchers were conducting authorized security testing, such as a penetration test or bug bounty exercise, or whether the access was obtained through other means. The identity of the researchers, the specific vulnerabilities exploited, and the scope of the data accessed were not detailed in the excerpt of the report reviewed.

Neither OpenAI nor Anthropic has issued a public statement addressed in the reviewed material confirming or responding to the report. This article will be updated if further details or statements from the companies involved become available.

The incident, as described, highlights ongoing concerns in the technology industry about the potential for advanced AI systems to be used as tools in cybersecurity contexts, including both offensive and defensive applications, though the precise circumstances of this case have not been independently verified beyond the initial report.