Unit 42: AI Agents Ran an Entire Ransomware Attack in 10 Hours, Then Left an 80-Page Audit

Palo Alto Networks' Unit 42 incident response team documented a ransomware intrusion in which a human attacker used frontier AI models and an agentic attack framework to autonomously dismantle an enterprise's cloud, identity, CI/CD, and SaaS defenses in roughly 10 hours — exploiting more than 50 distinct MITRE ATT&CK techniques along the way. Unit 42 estimates the same scope of work would normally take a coordinated human red team approximately two weeks.

How the AI Agents Divided the Work

• A dedicated reconnaissance agent automatically mapped the target company's internal microservices architecture
• Sub-agents combed through enterprise code repositories, extracting hard-coded tokens and service passwords embedded directly in source files
• A separate agent used those harvested credentials to infiltrate the organization's secrets management system and obtain root-level administrative control across the environment
• A pipeline-focused agent hijacked CI/CD workflows specifically to exfiltrate cloud access keys, completing the chain from initial recon to full environment compromise

The Detail That Stands Out: An AI-Written Audit

Beyond speed, the case is notable for what the attacker left behind: an 80-page security audit of the victim's environment, generated by the same AI agents used to carry out the breach, delivered to the victim during ransom negotiations. It is a small but telling illustration of how thoroughly agentic tooling now documents its own work — capability that, applied defensively instead, is exactly the kind of environment mapping most security teams struggle to produce for themselves.

What This Means for Enterprise Security Planning

The specific numbers here — 10 hours versus two weeks, 50-plus ATT&CK techniques chained autonomously — matter less than the underlying shift: attack timelines that used to give a defending team days or weeks of detection opportunity are compressing to hours. Hard-coded credentials in source repositories and CI/CD pipelines with excessive access are not new weaknesses, but this incident is a concrete demonstration that AI-assisted attackers now exploit that specific, common weakness pattern faster and more completely than a human team typically would.