Google detects first AI-developed zero-day exploit in the wild

In a landmark May 2026 report, Google's Threat Intelligence Group (GTIG) revealed the first confirmed instance of a threat actor developing a zero-day exploit using artificial intelligence. The exploit targeted a popular open-source, web-based system administration tool, bypassing its two-factor authentication mechanism.

The vulnerability itself was not a typical memory corruption or input sanitization flaw. Instead, it was a high-level semantic logic error — a hardcoded trust assumption in the 2FA implementation that traditional fuzzers and static analysis tools would not detect. This is precisely the type of vulnerability that frontier LLMs excel at finding: contextual reasoning gaps where the code appears functionally correct but is strategically broken from a security perspective.

GTIG's analysis of the exploit script found hallmarks of AI generation, including an abundance of educational docstrings, a hallucinated CVSS score, and a structured, textbook Pythonic format. The threat actor had partnered with others to plan a mass exploitation campaign, but Google's proactive discovery and responsible disclosure to the affected vendor disrupted the operation before it could be deployed at scale.

The implications extend beyond this single vulnerability. GTIG also documented state-sponsored actors from China (APT45) sending thousands of automated prompts to analyze CVEs and validate proof-of-concept exploits at scale. North Korean and PRC-nexus actors have integrated specialized vulnerability databases into AI workflows to prime models for security research. The era of AI-accelerated vulnerability discovery has arrived — for both attackers and defenders.