AI agents broke into 100+ company sites for one hacker

A security researcher says AI agents run for one Chinese-speaking hacker attempted more than 1,200 websites and got into sites belonging to more than 100 companies. The agents ran on Cairn, an open-source AI problem-solving platform, and did the work with little human input. Reuters, which reported the story on October 19, could not establish how much data was taken.

What happened

Snorre Fagerland, principal threat researcher at the security company NorForge, found Cairn control panels left exposed on the web. Through them he could watch the operators at work.

• The scale. The hacker's agents attempted to break into more than 1,200 sites, according to Reuters. They got into websites belonging to more than 100 companies.
• The targets. Reuters says the campaign went after e-commerce websites in the US, Canada, Australia and Japan. NorForge's own report says the group's targets were "mostly in the US, Japan and Australia."
• The platform. The agents ran on Cairn. Reuters calls it an open-source AI problem-solving platform. Its author calls it a multi-agent architecture.
• The instructions. Reuters reports that the operator's instructions were fewer than 100 Chinese characters long.
• The cost. Fagerland estimated that the campaign used roughly $5,000 worth of AI tokens, or about $50 per successful intrusion. These are researcher estimates. NorForge's published report gives a separate estimate of about $5 per task, counting failed attempts, and says the uncertainty is "considerable."
• The human role. Fagerland says the hacker did almost none of the work. He told Reuters his find showed commercially available AI agents "could pull off complex intrusions from start to finish without human intervention." The agents spent hours probing sites on their own. Successful intrusions took a little under 23 hours on average.
• The data. Reuters could not establish how much data, if any, was taken in the breaches.

NorForge's report did not name any victims. Reuters identified some of them by reviewing the hacker's Cairn activity. Three companies responded to Reuters:

• Deere said it had taken "necessary remediation actions" and had "no indication that sensitive information was accessed."
• SharkNinja said the hack hit a standalone site that carried only product support information, and that the threat had been contained.
• Cummins said it was investigating a report of a potential attack on a website run by one of its third-party vendors.

Two smaller businesses also acknowledged being compromised, Reuters said. Other victims did not reply.

Timeline

• September 2026 (day not given): Gambit Security reports a criminal campaign that used AI tools, including Cairn (NorForge, Reuters).
• September 25, 2026: The Chinese-speaking hacker's campaign begins (Reuters).
• Around October 2, 2026: Fagerland begins finding exposed Cairn instances used for malicious activity (Reuters).
• Late in the week before publication (date not given): The agents have attempted more than 1,200 sites (Reuters).
• Date not given: Deere, SharkNinja and Cummins give statements to Reuters (Reuters).
• October 19, 2026: NorForge publishes its report. Reuters publishes its exclusive, republished by CTV News at 9:58 a.m. EDT (NorForge, CTV News).

Confirmed vs not confirmed

What they said

• "It's a new world now." Snorre Fagerland, NorForge, to Reuters via CTV News
• "No indication that sensitive information was accessed." Deere, to Reuters via CTV News
• "Investigating the report of a potential cyber-attack on a website operated by one of our third-party vendors." Cummins, to Reuters via CTV News
• "If it is used to carry out harmful hacking attacks, that is not my intention." Cairn's author, on X, as reported by Reuters via CTV News
• "This could and should inform our defenses going forward." NorForge report

What Cairn is

Cairn is an open-source platform that coordinates several AI agents toward a goal. A user gives it a target and an objective. The agents then work step by step and share what they find with each other. NorForge describes it as "a state-based problem solving engine." It says the tool can be used for penetration testing, which is legitimate security testing done with permission.

Cairn's author says it was not built for hacking. NorForge also found real client penetration tests on unsecured Cairn panels, which it called "highly problematic."

What to check on your website this week

None of the named companies has said how its site was entered. These general checks reduce the risk from automated attacks.

Don't block the wrong bots

Not every automated visitor is an attacker. AI search engines and assistants also send crawlers and agents to your site, and those visits help your pages get cited in AI answers. Before you block traffic, separate known AI crawlers from unknown automated traffic in your logs. If you want to see whether AI search engines actually cite your brand, a tool like Xtrusio can measure your AI visibility. It is not a security product, so keep your security checks with your IT team and hosting provider.

Sources

All sources read on October 20, 2026.

• Reuters exclusive by Raphael Satter, republished by CTV News, October 19, 2026: How one hacker used AI agents and a few thousand dollars to break into scores of sites
• NorForge, by Snorre (Fagerland), October 19, 2026: Inside Cairn: Open Windows into massive AI-Driven Cyber Operations
• TNW, by Ana Maria Constantin, October 19, 2026: AI agents broke into 100+ sites for one hacker at $50 each, researcher says