Most stories about AI and cyber security in 2025 were about attackers. This one was the other way round. On 15 July 2025 Google said that its AI agent Big Sleep had found a critical software flaw that criminals already knew about and were preparing to use, and that it was fixed before they could.

What Big Sleep is

Big Sleep is an AI agent developed by Google DeepMind and Google Project Zero, Google's elite vulnerability research team. It reads and tests software looking for security flaws, the way a human researcher would. By November 2024 it had found its first real-world vulnerability, an unreleased bug in SQLite.

The SQLite flaw

SQLite is a small database engine built into phones, browsers, apps and countless other products — one of the most widely deployed pieces of software in the world.

According to Google, "based on intel from Google Threat Intelligence, the Big Sleep agent discovered an SQLite vulnerability" — "a critical security flaw, and one that was known only to threat actors and was at risk of being exploited."

SQLite's own description of the flaw, now tracked as CVE-2025-6965, is that "An attacker who can inject arbitrary SQL statements into an application might be able to cause an integer overflow resulting in a read off the end of an array." It was fixed on 27 June 2025, in SQLite 3.50.2.

Google's claim: "We believe this is the first time an AI agent has been used to directly foil efforts to exploit a vulnerability" in the wild. It did not name the threat actors or share technical indicators (The Record, SecurityWeek).

Other AI tools for defenders

In the same announcement Google highlighted other AI-assisted security tools:

  • Timesketch, its open-source investigation platform, which "will accelerate incident response by using AI";
  • FACADE, a system Google has used internally since 2018 to detect insider threats;
  • support for the AI Cyber Challenge at DEF CON 33 and a white paper on building secure AI agents.

Why it matters

Attackers are using AI to find and exploit flaws faster. Big Sleep is an example of the same technology helping defenders find flaws first. Speed matters on both sides: the earlier a flaw is fixed, the smaller the window for attackers.

What it means for you

At home: you will rarely hear about fixes like this; they arrive silently through app and system updates. That is why keeping automatic updates on matters.

If you run a small business: AI-powered security is becoming a standard feature of good security products. Ask suppliers how they use it — but judge them on results like patch speed and detection, not on marketing claims.

If you work in IT or at an MSP: expect more vulnerabilities to be discovered faster, by both attackers and AI-assisted researchers. That means more patches, sooner. Make sure your patching process can keep up, and consider AI-assisted tools for triage and investigation.

The bottom line

Big Sleep's SQLite find was a small but significant moment: proof that AI can help defenders win a race against attackers. The practical lesson is unchanged — the benefit only reaches you if the fix is installed.

This article summarises and comments on the sources listed below. It is general information, not legal or professional advice.