AI-Assisted Ransomware Gang Tore Through a Corporate Network in Under 10 Hours

Key points

• Palo Alto Networks' Unit 42 research team documented a ransomware intrusion completed in under 10 hours that analysts estimated would take human operators roughly two weeks without AI assistance.
• The attackers used AI agents, software programs that act independently and adjust their own next steps based on what they find, to automate credential theft and network reconnaissance.
• The criminals broke in through a public-facing API endpoint, a doorway that lets outside software talk to internal systems, before pivoting to cloud services and a secrets-management vault holding administrator passwords.
• Existing branch protections, a setting that requires human approval before code changes go live, blocked the attempt to plant hidden backdoors inside infrastructure-management files.
• The core risk isn't new attack techniques but a sharp drop in the time and cost needed to run existing ones at speed.

A ransomware gang used AI to move through a corporate network in fewer than ten hours, according to Unit 42, the threat-research arm of Palo Alto Networks. Ransomware is malicious software that locks a company's files until a payment is made. Researchers put the equivalent human-only effort at around two weeks.

The group has not been publicly named. What Unit 42 documented, first reported by CSO Online, is a playbook built on familiar criminal techniques handed to AI agents: software that reads each step's result and decides the next move without waiting for a human instruction. Unit 42 counted more than 50 techniques across the intrusion, mapped against the MITRE ATT&CK framework, a standard catalogue of attacker behaviour.

We've covered Palo Alto Networks' threat research frequently this quarter. In August, our analysis of 405 AI-linked malware samples found that existing defences caught every one that reached a real computer. This incident is different: the AI wasn't the weapon, it was the operator.

How did the attackers get in?

Entry came through a public-facing API endpoint, a connection point that lets external software communicate with internal systems. An automated reconnaissance agent then mapped the internal network, identifying services much like a burglar casing each room.

Separate agents searched source-code repositories, the libraries where developers store their work, for exposed credentials left in the code by mistake. Those credentials unlocked a secrets-management vault where administrator passwords and access keys were stored. Vault access meant full administrative control.

The attackers hijacked a code-management application to steal cloud access keys, digital passes that grant access to rented computing resources. They turned the victim's own cloud power into infrastructure for further attacks. Their bid to embed backdoors inside Terraform configuration files, written instructions companies use to build cloud infrastructure, was stopped by branch protections requiring human sign-off on code changes.

During ransom negotiations the attackers confirmed to Unit 42 that they'd used frontier AI models and purpose-built attack frameworks. Whether the attack was fully autonomous isn't settled. One analyst described it as a human-directed intrusion in which AI handled the tactical, step-by-step execution.

Should businesses be worried about the speed?

Yes, because most security teams aren't built to respond in hours. Detection-and-containment processes were designed for intrusions that unfold over days.

Advisors on the case recommend moving away from long-lived credentials, passwords and keys that stay valid for months, toward short-lived credentials that expire quickly. Security providers should also be authorised in advance to disable a breached account without waiting for internal approval, shrinking the window between detection and containment.

This attack also illustrates how difficult it is to spot malicious activity that crosses systems monitored separately. A suspicious action in one environment looked unremarkable alone; its significance only became clear when connected to activity elsewhere. That's a structural problem, not a configuration tweak.

What should ordinary employees watch for?

This intrusion targeted technical infrastructure rather than individual inboxes. Stolen credentials are still the most common entry point across ransomware cases, though. Employees who reuse passwords across work and personal accounts, or store them in shared documents or code files, create exactly the kind of exposure exploited here. Our earlier story on the Passportal vault breach showed how quickly that access can be drained once it's found.

If your company notifies you of a security incident, change your work passwords straight away and watch for unexpected password-reset emails you didn't request.