Google’s Threat Intelligence Group (GTIG) has uncovered a troubling new trend: hackers are now using AI large language model (LLM) frameworks to automate the discovery and exploitation of software vulnerabilities without human intervention. In a report released Monday, GTIG detailed a tool found on an exposed staging server that could compare software versions, identify flaws, and generate exploit code autonomously.
The researchers noted that in the past quarter, they observed threat actors deploying fully autonomous pipelines to exploit vulnerabilities. This marks the first time such automation has been identified in the wild. The tool was set up to conduct continuous differential analysis, comparing code before and after patch announcements on popular web applications, enabling it to autonomously discover and exploit vulnerabilities.
Google warned that these automated workflows significantly reduce the time organizations have to apply security patches before attackers develop exploits. The report highlighted that vulnerability disclosures more than doubled between January and August 2026, with the number of zero-day vulnerabilities exploited averaging 11 per month this year, peaking at 22 in August.
Anthropic, another AI company, also reported that AI-assisted attacks have evolved into multi-agent operations, automating most of the cyberattack cycle from discovery to data theft. This evolution allows threat actors to run multi-victim campaigns that would previously have required teams of operators.