Threat researchCisent
One attacker set AI agents loose to hack on their own, and they broke into the websites of over 100 companies
A researcher watched AI agents explore and break into websites for hours, with almost no human input. That way, a single attacker got into companies such as Deere, SharkNinja and Cummins.

On October 19, 2026, Reuters published what Snorre Fagerland, principal threat researcher at the security firm NorForge, watched for two weeks: a Chinese-speaking attacker who broke into websites while doing almost nothing. The work was done by AI agents that spent hours exploring and attacking on their own.1
What happened
Fagerland was scanning the internet for exposed installations of Cairn, an open-source AI agent platform. From October 2 he found about two dozen being used to attack, and through them he could watch the agents at work.1
Starting September 25, this attacker’s agents tried to break into more than 1,200 e-commerce sites in the United States, Canada, Australia and Japan, and succeeded with the websites of more than 100 companies. Reuters identified Deere, SharkNinja and Cummins among them. The attacker’s instructions took up fewer than a hundred Chinese characters, and each successful intrusion took about 23 hours of exploring and testing. Fagerland estimated the campaign cost about US$ 5,000 in AI model usage, roughly US$ 50 per successful attack.1
In one case, a large company’s main application had no weaknesses. So the agents looked for its associated infrastructure, found an admin panel exposed to the internet, uncovered a subtle flaw in how it processed data and, after hours of trial and error, got the server to run their commands.1
Deere said it took the necessary corrective measures and sees no sign that sensitive information was accessed. SharkNinja said the attack hit a separate support site and has been contained. Cummins said it is investigating a possible attack on a site run by one of its suppliers. Reuters could not determine how much data was taken.1
Why it matters
An intrusion that used to take hours of work by someone experienced is now carried out by an agent on its own, for a few dozen dollars, and the same attacker repeats it hundreds of times.1 For Fagerland, it is proof that commercially available agents can now carry out complex intrusions from start to finish without human intervention.1
And the way in was not the main application, but a forgotten panel on a neighboring server. Agents go through everything an organization has exposed, not just what it looks after.
Where Cisent comes in
Sentinel Red Team attacks the organization’s systems with its authorization and within the scope it defines, to find the gaps before an attacker does: the admin panel exposed on a server nobody watches, the service published by mistake, the insecure configuration. It’s ongoing testing, not one audit a year, and every new attack pattern helps train the defense.
Sources
Reuters, “Exclusive-How one hacker used AI agents and a few thousand dollars to break into scores of sites”, October 19, 2026. https://money.usnews.com/investing/news/articles/2026-10-19/exclusive-how-one-hacker-used-ai-agents-and-a-few-thousand-dollars-to-break-into-scores-of-sites · Spanish version in Diario Financiero. https://www.df.cl/df-lab/mundo/como-un-hacker-utilizo-agentes-de-ia-y-unos-pocos-miles-de-dolares-para↩︎