Hacker Suspected of Exploiting Anthropic Mythos to Breach GitHub, Copilot Source Code Exposed

According to Revelio Intelligence monitoring, the SlowMist security team revealed that the latest disclosure on a dark web forum indicated that the attacker allegedly used Anthropic's security AI model "Mythos" to precisely breach GitHub's internal defenses.

Within the stolen approximately 4000 GitHub internal code repositories, the leaked data includes the source code for Copilot, the CodeQL algorithm, Actions runtime, and the billing system's code. The SlowMist team warned that once hackers deeply analyze this code, it could have subsequent security implications for the open-source community. The reason is simple: these systems underpin many developers' core workflows.