Google has reached a new milestone in the cybersecurity world with its artificial intelligence agent, Big Sleep. Developed by DeepMind in collaboration with Google’s Project Zero team, Big Sleep recently detected and autonomously blocked a real-time cyber exploit before it could be executed. This marks the first recorded instance of an AI system preventing a live attack entirely on its own, showcasing a transformative step in the evolution of cybersecurity defense strategies.

The attack in question targeted a previously unknown vulnerability in SQLite, a widely used open-source database embedded in countless applications and devices. Known officially as CVE‑2025‑6965, this vulnerability had not been disclosed to the public and was only known to threat actors. Big Sleep’s detection and intervention demonstrate the power of predictive, intelligent systems that go beyond traditional threat scanning to anticipate and neutralize attacks as they emerge in real time.

What makes Big Sleep remarkable is its ability to analyze massive amounts of threat data, learn from past patterns, and apply this intelligence to new scenarios without human instruction. This incident highlights how AI can not only detect vulnerabilities but also act on them before they’re exploited. It represents a shift from reactive security—where systems respond after an attack occurs—to proactive defense, where AI works autonomously to protect digital environments.

This achievement comes as part of Google’s broader initiative to integrate AI into its security platforms. Other tools, such as Timesketch powered by the Sec-Gemini model, and FACADE, which detects insider threats, are also part of Google’s push to create smarter, more anticipatory security infrastructure. Big Sleep’s success further validates Google’s approach to developing autonomous cybersecurity systems capable of adapting to evolving threat landscapes.

Google has already started integrating Big Sleep’s capabilities into more areas of its cloud and infrastructure, with plans to expand its use in safeguarding critical open-source projects. With millions of users depending on secure digital platforms, such innovations in AI promise to enhance not just Google’s products but the global security posture at large. This moment could very well mark the dawn of AI-driven autonomous cyber defense.

Sources:

• Google’s Big Sleep AI foils live zero-day exploit in SQLite
  https://securityonline.info/googles-big-sleep-ai-foils-live-zero-day-exploit-in-sqlite-cve-2025-6965/
• Google’s Big Sleep blocks cyberattack – Sundar Pichai confirms
  https://www.livemint.com/technology/googles-ai-agent-big-sleep-foils-cyberattack-in-groundbreaking-first-says-sundar-pichai-11752602040494.html
• How Big Sleep became Google’s secret cyber watchdog
  https://economictimes.indiatimes.com/news/new-updates/how-big-sleep-became-googles-secret-cyber-watchdog/articleshow/122577437.cms
• Google is doubling down on cybersecurity using AI
  https://www.androidcentral.com/apps-software/google-is-doubling-down-on-cybersecurity-using-ai
• Vice: Big Sleep becomes first AI to autonomously prevent cyberattack
  https://www.vice.com/en/article/google-big-sleep-first-ai-to-ever-prevent-cyberattack