Google's Threat Intelligence Group (GTIG) said it has identified attackers using AI large language model frameworks to automate software vulnerability discovery and exploitation for the first time, according to a report published Monday.
Researchers described a tool found on an exposed staging server that compared software versions, identified potential flaws, and generated and executed exploit code. The multi-agent framework conducted continuous differential analysis comparing code before and after patch announcements on popular web applications, autonomously discovering and exploiting vulnerabilities.
Compressed Patching Windows
The automated workflows significantly reduce the time organizations have to apply security patches before attackers develop exploitation methods. GTIG warned that AI agent frameworks are enabling attackers to scale exploitation pipelines against internet-facing servers without human intervention.
Vulnerability disclosures more than than doubled between January and August, according to Google. The number of zero-day vulnerabilities exploited averaged eight per month in 2025 and 11 per month from January through August 2026, reaching 22 in August.
Accelerated Attack Cycles
These automated workflows also accelerate attackers' ability to move from initial access to gaining domain administrator access. In an August 2026 PaperCut campaign, attackers achieved administrative control within minutes, according to cybersecurity firm GreyNoise.
Cloudflare reported attempts to exploit PHP flaws against its customers in October. GTIG said that technical information published before the October 21 patch enabled researchers and attackers to study vulnerabilities and attempt exploits using LLM analysis.
AI company Anthropic also warned that AI-assisted attacks have evolved into multi-agent operations automating most of the cyberattack cycle from discovery to exploitation and data theft. Anthropic described threat actors running multi-victim campaigns using agentic AI that would previously have required teams of operators.


