According to reports, two weeks after a group of AI agents broke through the blockade and attacked Hugging Face from the OpenAI system, the ChatGPT developer discovered another intrusion driven by artificial intelligence — and this time, they themselves were targeted. Using Claude software developed by Anthropic, a team of independent security researchers successfully obtained the ChatGPT account permission of an OpenAI employee, thus being able to read and modify the content of the company's private software cache. The team participated in OpenAI's vulnerability mining program, which provides a secure environment for researchers to attempt to hack enterprise systems. Once they discovered the vulnerability, they quickly reported it to the company. OpenAI paid a $6,500 reward for this, and the team also revealed the details of the operation to the media for the first time. OpenAI said the hackers discovered two issues: one was a bug in Discourse, a third-party service hosting the OpenAI community discussion forum, and the other was an issue with the company itself. According to OpenAI, these two issues have now been resolved. “We are grateful to the researchers for reaching out to us and sharing their findings,” the company said. We've reduced the scope of community login tokens and revoked affected tokens and sessions.” An Anthropic spokesperson declined to comment.

According to reports, two weeks after a group of AI agents broke through the blockade and attacked Hugging Face from the OpenAI system, the ChatGPT developer discovered another intrusion driven by artificial intelligence — and this time, they themselves were targeted. Using Claude software developed by Anthropic, a team of independent security researchers successfully obtained the ChatGPT account permission of an OpenAI employee, thus being able to read and modify the content of the company's private software cache. The team participated in OpenAI's vulnerability mining program, which provides a secure environment for researchers to attempt to hack enterprise systems. Once they discovered the vulnerability, they quickly reported it to the company. OpenAI paid a $6,500 reward for this, and the team also revealed the details of the operation to the media for the first time. OpenAI said the hackers discovered two issues: one was a bug in Discourse, a third-party service hosting the OpenAI community discussion forum, and the other was an issue with the company itself. According to OpenAI, these two issues have now been resolved. “We are grateful to the researchers for reaching out to us and sharing their findings,” the company said. We've reduced the scope of community login tokens and revoked affected tokens and sessions.” An Anthropic spokesperson declined to comment.