Google has officially detected the first zero-day exploit created by artificial intelligence, marking a significant milestone in automated cyber threats. The vulnerability was designed to bypass two-factor authentication (2FA) within a web-based system administration tool, though security measures successfully blocked the attack before it could be deployed in a live environment.
The threat actors utilized an AI model to analyze the structure and logic of Python code to pinpoint the flaw. Interestingly, the exploit script contained evidence of AI hallucinations, confirming the code was generated by a large language model rather than a human developer. While Google has ruled out its own Gemini model as the source, the specific AI used to craft the attack remains unidentified.
