Hacker uses AI agents for entire attack chain, mocks victim with detailed security audit
AI agents reduced the mission time from 2 weeks to less than 10 hours.
AI agent is a hacker's best new friend. By Cybernews.
- Unit 42 says a ransomware attacker used AI agents to breach an enterprise in under 10 hours.
- The agents mapped services, stole tokens and passwords, and helped reach cloud and identity systems.
- The attacker left an 80-page security audit that listed dozens of exploited weaknesses.
- Unit 42 says defenders need automated responses that can revoke access and isolate cloud accounts quickly.
Key Takeaways by nexos.ai, reviewed by Cybernews staff.
Not only did a human ransomware attacker breach an enterprise network in less than 10 hours with the help of AI agents, but they also left an 80-page security audit for the victim company.
According to Unit 42, a security arm of Palo Alto Networks, the human hacker told ransom negotiators that they used frontier AI models – they’re unnamed – and agentic attack frameworks with AI agents carrying out each step of the breach.
After responding to the incident, Unit 42 soon established that the AI agents reached the company’s security layers in a methodical manner, each targeting a different layer of defense to achieve a shared goal.
The attack was thorough but was executed extraordinarily quickly. Human operators would normally take around two weeks to complete the breach, but the attacker “compressed weeks of methodical intrusion tradecraft” into less than 10 hours.
What made the attack stand out was AI-assisted operational efficiency, without the need for a novel zero-day or super elite tradecraft,Unit 42 said in their write-up of the investigation.
“The attacker left tactical execution to AI agents that monitored, evaluated, acted, and re-planned in real time, increasing speed throughout the attack chain.”
First, the AI agents performed reconnaissance and breached a public API endpoint to gain access to the enterprise network.
Then, the attacker sent in an automated recon agent to map internal microservices. Simultaneously, extra subagents scraped code repositories to pilfer hard-coded tokens and service passwords.
Those were used to access the organization’s secret management system and steal master credentials to gain access to the root system.
Has your password leaked?
“Specialist pivot agents” then validated access to cloud, identity, CI/CD container, and SaaS environments. CI/CD workflows were hijacked to steal cloud access keys and turn the organization’s cloud AI services into post-compromise infrastructure.
“This allows threat actors to hide orchestration traffic among expected traffic, and offload the financial cost onto the victim,” Unit 42 pointed out.
The attacker also used a “documentation agent” to prepare a detailed security audit for the successfully targeted company and leave it behind on the organization’s security posture. The “report” details dozens of exploited findings.
Stay updated with our latest stories and follow us on social media
Be the first to discover new stories, ideas, and updates from our team.
According to Unit 42, defending against automated agent loops requires matching the speed and adaptability of AI-driven attacks.
In other words, defenders should also deploy automated playbooks “that simultaneously revoke credentials, terminate OAuth sessions, freeze CI/CD pipelines, and isolate cloud accounts across all operational planes.”
AI agents are becoming a significant issue in the world of cybersecurity. Just recently, researchers found that hackers are leveraging AI agents en masse to make them install malicious code on company websites.
And in August, another group of researchers said AI agents had carried out a 4-day cyberattack against Taiwanese government agencies with minimal human input. The agents mapped networks, found weaknesses, exploited systems, and collected data.
