Google’s Project Zero hackers and DeepMind boffins have collaborated to uncover a zero-day security vulnerability in real-world code for the first time using AI.

The flaw, an exploitable stack buffer underflow in SQLite, was found by Google’s Big Sleep team using a large language model (LLM)

Google Project Zero and Deep Mind informed SQLite maintainers in October, who were able to devise a fix on the same day.

Google's Big Sleep AI framework uncovers a zero-day vulnerability in SQLite, showcasing AI-driven security potential.

Google's "highly experimental" artificial intelligence agent Big Sleep has autonomously discovered an exploitable memory flaw in popular open-source