Google confirms first fully automated AI cyberattack

news.mynavi.jp (Japanese) —

Google has confirmed the first instance of AI executing a cyberattack without human intervention, automating everything from vulnerability identification to code creation and execution. The framework, called "exploit_agent," was discovered on a server and uses multiple large language models to compare source code before and after security patches, identifying vulnerabilities and launching attacks automatically. This reduces preparation time, potentially shortening the window for companies to apply updates. Google warns that AI enables faster exploitation of known "N-day" vulnerabilities, citing an August 2026 attack on PaperCut software. While AI attacks may leave detectable traces, the report highlights growing automation tools like ARTEX and CAIRN, urging companies to accelerate patch deployment.


With a significance score of 5.8, this news ranks in the top 0.5% of today's 32692 analyzed articles.

Get summaries of news with significance over 5.5 (usually ~10 stories per week). Read by 10,000+ subscribers: