A hacker reportedly ran AI agents against scores of sites for a few thousand dollars

One hacker used AI agents and only a few thousand dollars to break into scores of websites, according to an exclusive from Reuters.

Summaries of the report, including one from Techmeme, say the account comes from a researcher. They describe the hacker as Chinese-speaking and list Deere and SharkNinja among the affected organizations. The Verge also covered the story, under a headline about a hacker who broke into John Deere and SharkNinja with an army of AI agents.

We could not open the full Reuters text. This report rests on its public headline and on those summaries. Every claim here is the reporting’s, not something we have confirmed.

A small budget behind a wide campaign

The central claim is about cost. In its headline, Reuters says the hacker needed only a few thousand dollars to get into scores of sites. AI agents are software that can plan and carry out steps with little human input, so one person can run many intrusion attempts at once. That is the mechanism the headline points to.

Reuters calls the hacker a single person. Neither the headline nor the summaries we read name that person or link the activity to any government. Any such link would be speculation.

What the public reporting leaves open

Several basic questions are not answered in the text available to us:

• What data, if any, was taken from the companies named.
• Which AI models powered the agents.
• How the researcher tracked the activity.
• How the companies responded.

We have not seen statements from Deere or SharkNinja, so we do not describe their response. The companies are named in the reporting as affected organizations, which is not the same as confirmation from them.

What would confirm or weaken the account

This is a single-source story, so follow-up matters. Three developments would firm it up or undercut it:

• Technical findings published by the researcher.
• Statements from the affected companies on what was accessed.
• Action by the AI providers whose services were allegedly used.

The last one would show whether model makers can spot this kind of automated activity. For security teams, the practical point in the headline is the price. If a few thousand dollars can reach scores of sites, defenders face attackers who can try far more targets than before. Basics such as patching, access limits and monitoring on public-facing sites become more urgent. We will update this report if Reuters or the companies publish more details.