One Hacker Used AI Agents to Breach More Than 100 Company Websites

A single hacker used commercially available AI agents to break into the websites of more than 100 companies, according to a Reuters exclusive published on October 19, 2026. The victims included the equipment makers Deere and Cummins and the appliance company SharkNinja. The campaign ran on a budget of a few thousand dollars in AI tokens. The reporting was carried by BNN Bloomberg and CTV News, and the analysis came from Snorre Fagerland, a principal threat researcher at NorForge.

Reuters says the AI agents attempted to breach more than 1,200 sites in total, and that a successful intrusion typically took about 23 hours. The researcher estimated the campaign used roughly $5,000 in tokens, or about $50 per successful hack. A separate NorForge estimate, as summarised by The Next Web, puts the average at about $5 per task, counting failed attempts, and says the uncertainty is considerable. The two figures measure different things, so neither should be read as a final cost.

Several AI tools were involved. Reuters says the campaign appeared to use a mix of DeepSeek’s models and Anthropic’s Claude Code, along with Cairn, an open-source AI problem-solving platform that the researcher found exposed online. The Next Web reports that NorForge suggests the agent may have been Claude Code running a DeepSeek V4.1 Flash model. The reporting says neither Anthropic nor DeepSeek responded to Reuters’ requests for comment.

Fagerland told Reuters that the case showed commercially available agents could carry out complex intrusions without human intervention. He described the techniques as “not script kiddie stuff” and said the pace of development was astounding. That framing matters because the attack did not need specialist skills at each step. The agents handled the routine work of finding and exploiting weak points, which lowered the cost of reaching many targets at once.

The affected companies responded differently. Deere said it had taken “necessary remediation actions” and had no indication that sensitive information was accessed. SharkNinja said the intrusion hit a standalone site that held only product support information, and that the threat was contained. Cummins said it was investigating a potential attack on a website run by a third-party vendor. Two smaller businesses acknowledged they had been compromised, and other victims did not respond.

The developer behind Cairn acknowledged Reuters’ message but did not answer its questions. In a post on X, that person said the platform was not designed for hacking and that harmful use was not the intention. Whether a platform’s stated purpose limits what its users can do is a question that the case raises without answering.

What is still unknown is who the hacker is, which companies make up the full list of 100-plus victims, and what data, if any, was taken. The reporting does not list every affected company, and the cost estimates differ by a factor of ten. Site owners who want to reduce the risk of credential abuse can start with unique, long passwords, which the free password generator at /en/password-generator can produce.