A rivalry between competing AI companies, perhaps. According to Hacktron, three security researchers using Anthropic's renowned Claude AI were able to impressively breach the internal systems of top-tier AI company OpenAI within 72 hours. During an authorized security test conducted in July 2026, the researchers first discovered a vulnerability in OpenAI's community forum running on the Discourse platform.
Claude was the key player in this operation. With the assistance of Claude Opus 4.8 and the newly released Claude Opus 5, the researchers were able to develop a method to fully control the forum server within hours. Subsequently, when they connected to another vulnerability in OpenAI's single sign-on (SSO) system, they gained direct access not only to employee secret accounts but also to OpenAI's most critical private GitHub repository.
However, since this was purely a security test, the researchers only created a harmless pull request to demonstrate their access and did not attempt to steal sensitive information. Upon discovering this cyber breach, OpenAI patched the vulnerabilities within 14 hours and rewarded the researchers with a bug bounty of $6,500 for identifying the vulnerability.
Ref: Technology Innovation
