A trio of security researchers successfully hacked into OpenAI employee accounts in under 72 hours, utilizing Anthropic's Claude Opus 4.8 and 5. They accessed OpenAI's GitHub repository, known as 'Monorepo,' which is believed to contain sensitive algorithmic information. The team exploited a vulnerability in Discourse, the platform hosting OpenAI's forums, to gain access.