AI agents have carried out an entire ransomware attack chain in under ten hours, according to a case study from Palo Alto Networks' Unit 42.

The September 2 case shows the agents handling every stage on their own: reconnaissance, intrusion, lateral movement and encryption. They even left behind an 80-page audit of their own work.

Researchers say the case is a warning about how fast offensive AI capabilities are advancing. Tasks that once needed skilled human operators can now be automated.

The agents did not need novel exploits. They used known techniques, chained together quickly and adapted when defences blocked them.

Security teams say the finding raises the stakes for patching and monitoring. Slower defenders will struggle against machine-speed attackers.

Unit 42 urged organisations to assume AI-assisted attacks are already happening and to test their defences accordingly.

Ransomware gangs have increasingly experimented with AI for phishing emails and malware development. Cyber insurers say AI-assisted attacks are already pushing premiums higher for poorly defended firms.

Sources: tech-insider.org, stackfutures.com, cybernews.com.