A team of US based startup cyber security researchers successfully conducted a hack against OpenAI, utilizing assistance from Anthropic’s Claude chatbot. This operation resulted in the compromise of a number of ChatGPT accounts belonging to OpenAI employees. The researchers were able to initiate a process through these compromised accounts, granting them initial access.
This initial access enabled the team to reach the targets’ software cache, with indications that potentially more systems could have been compromised. The cyber security researchers involved declared that the theoretical scope of what they could access was significant. This incident represents the latest in a series of security issues faced by OpenAI, highlighting continued vulnerabilities for the company.