In May 2026, the Google Threat Intelligence Group (GTIG) confirmed the first known instance of a cyberattack where artificial intelligence was used to develop and weaponize a zero-day exploit for a real-world campaign. The exploit targeted an open-source web administration platform to bypass multi-factor authentication, though the vendor patched the vulnerability before a mass exploitation event could occur.
How the AI-Generated Exploit Was Identified
Researchers identified the exploit as machine-generated by analyzing specific anomalies in the code, according to GTIG’s report. The Python-based script contained "overly explanatory comments" and a "hallucinated CVSS score," which are not typical of human-authored malicious code.
Google noted that the script followed a structured, "textbook Pythonic format" that closely mirrors the training data often found in large language models (LLMs). While the specific cybercrime group and the targeted software remain unnamed, the company confirmed the exploit was designed to facilitate a mass attack by leveraging valid credentials alongside the newly discovered flaw.
The Role of AI in Vulnerability Discovery
The incident suggests that AI is currently being used to accelerate existing workflows rather than inventing entirely new attack techniques. According to GTIG, the technology helps threat actors discover and weaponize vulnerabilities more efficiently.
While Google stated they do not believe their own AI model, Gemini, was used to create this specific exploit, they maintain high confidence that the attackers used some form of AI model to support the discovery process. Beyond this specific incident, Google has observed that state-sponsored threat actors—specifically those linked to China and North Korea—are increasingly interested in using AI for these purposes. For example, some China-linked groups have already been observed deploying agentic tools like Strix and Hexstrike to target Japanese and East Asian technology firms.
Recommended Security Defenses
To counter the rise of AI-supported exploitation, security experts recommend shifting focus toward foundational defense mechanisms. According to IANS faculty, organizations should prioritize the following:
- Reduce Exposure: Limit the number of internet-facing systems to reduce the overall attack surface.
- Asset Visibility: Improve the ability to track and manage all assets within a network.
- Patch Management: Accelerate patch response processes to close vulnerabilities before they can be weaponized.
- Detection and Recovery: Strengthen capabilities to detect and recover from zero-day scenarios rapidly.
John Hultquist, chief analyst at GTIG, noted that the industry should view this incident as "the tip of the iceberg" regarding the future of AI in cybercrime. As AI tools become more accessible, the barrier to discovering and exploiting critical software flaws continues to drop, making proactive security posture essential for all technology providers.