Google has announced the discovery of an AI tool named 'exploit_agent' that autonomously discovers and exploits vulnerabilities without human intervention. The tool was found on an exposed server and is capable of monitoring software updates, identifying vulnerabilities, and launching attacks.
The primary targets of these attacks have been identified as French websites in the media, entertainment, and education sectors. This development highlights a growing trend towards more sophisticated and automated attack frameworks, particularly among threat actors in Asia.
The report from Google noted a significant increase in zero-day exploits, with 22 reported in August 2026 alone. This marks a substantial rise from the previous average of 11 zero-day exploits per month earlier in the year.
The overall number of vulnerability disclosures and exploits has nearly doubled between January and August 2026, indicating an escalation in cyber threats. The use of AI in these attacks represents a new level of sophistication in cybercrime.
Unknown threat actors are suspected to be behind these attacks, utilizing the 'exploit_agent' tool to target specific sectors and regions. The automation of these processes suggests a shift in how cyber threats are being executed.
