What’s new?
Google's Threat Intelligence Group (GTIG) says it has for the first time observed attackers finding and exploiting vulnerabilities fully automatically, with no human involved. In its GTIG AI Threat Tracker, the team describes an open server running an AI-powered framework called "exploit_agent" that integrates several language models.
The system compares software code before and after patches and derives vulnerabilities from the difference. It creates exploits and runs them on its own. Russian-language comments in the configuration files point to the operator's background. Data on the server suggests the targets were mainly French websites in media, entertainment and education.
Google places the current situation at level 5, where AI-powered tooling carries out several attack phases with minimal human involvement. Even stronger attacks are only months away. GTIG classes the development as a critical turning point. For now, such attacks gain speed at the expense of stealth. They leave conspicuous forensic traces that make detection easier.
What changes for companies?
- IT operations and patch management
- If attackers analyse patches automatically, every delay in applying them becomes a risk.
- Security monitoring
- Autonomous attacks leave conspicuous traces, so it is worth reviewing logging and detection.
- Risk assessment and management
- Google expects stronger attacks within months, so the topic belongs on the agenda now.
Check how quickly patches go live in your environment and whether your systems reliably flag conspicuous attack traces.
Our verdict
Verdict: ActAct now: measure how long patches take to reach production and shorten that time. Check whether your monitoring detects conspicuous attack traces. Put the topic on the agenda with your IT provider or security lead.
Assessment by digitice.io, kept separate from the news. We report on products neutrally, without purchase recommendations.