Security researchers hacked OpenAI using Anthropic’s Claude

A three-person cybersecurity research team used Anthropic’s Claude Opus 5 to exploit vulnerabilities in an OpenAI community forum, take over employee accounts and demonstrate access to the company’s private code repository.

The operation began on July 23 and was carried out by researchers from Hacktron AI, who disclosed the vulnerabilities to OpenAI and stopped testing without reviewing the company’s source code.

The Wall Street Journal, which interviewed the researchers and first reported the incident, said OpenAI paid the team $6,500 for the discovery.

The researchers reported the vulnerability to OpenAI through the company’s bug bounty program. However, testing of the third-party forum software itself was outside the scope of OpenAI’s bounty program.

The researchers, Harsh Jaiswal, Mohan Pedhapati and Rahul Maini, published a detailed technical write-up explaining how they combined a flaw in the software behind OpenAI’s community forum with a separate issue in the company’s login system.

The attack began at community.openai.com, a help forum powered by the third-party discussion platform Discourse.

Hacktron found that some uploaded image formats were processed through ImageMagick and a vulnerable version of the libheif image-decoding library. The vulnerability allowed specially crafted image data to trigger remote code execution, meaning an attacker could run commands on the forum server.