One hacker. An army of artificial intelligence agents. More than 1,200 websites targeted. What investigators uncovered should make businesses, governments, and ordinary internet users seriously question whether cybersecurity is prepared for what’s coming.
The Hacker Didn’t Need an Army. Artificial Intelligence Became One.
For years, the public has been warned about artificial intelligence replacing jobs, manipulating information, and reshaping the global economy.
But another transformation is unfolding in the shadows of the internet.
Artificial intelligence is beginning to change who can launch sophisticated cyberattacks, how much those attacks cost, and how little human involvement they require.
A Reuters investigation published October 19, 2026, describes a Chinese-speaking hacker who allegedly used autonomous AI agents to compromise websites associated with more than 100 companies.
The reported cost of the operation? Approximately $5,000 in AI processing expenses.
That figure deserves attention.
Not because cybercrime was previously impossible for individuals with limited resources, but because the technology appears capable of performing work that once demanded considerable technical expertise, patience, and direct human involvement.
The unsettling question is no longer simply whether hackers can use AI.
______________________________________________
🔴 Help keep this independent voice alive and uncensored.
Support independent journalism and help keep Chris Wick News going.
👉 Support Chris Wick on GoFundMe
It’s how many attacks one person can launch when the machines do most of the work.
More Than 1,200 Websites Targeted
According to Reuters, cybersecurity researcher Snorre Fagerland, principal threat researcher at NorForge, discovered the activity while examining publicly exposed installations of Cairn, an open-source AI problem-solving platform.
Beginning in early October, Fagerland identified approximately two dozen instances of the platform apparently being used for malicious purposes.
One operator stood out.
The Chinese-speaking hacker had reportedly instructed AI agents to target e-commerce websites across the United States, Canada, Australia, and Japan.
By late last week, those agents had attempted intrusions against more than 1,200 websites.
More than 100 company websites were reportedly compromised.
Among the companies whose websites were identified in the investigation were Deere & Company, SharkNinja and Cummins.
The company responses are important.
Deere said it had undertaken remediation and had no indication that sensitive information was accessed. SharkNinja said the affected website contained product support information and that the incident was contained. Cummins said it was investigating a possible attack involving a third-party vendor’s website.
Reuters could not establish how much information, if any, was extracted from the affected systems.
A compromised website is not automatically evidence that customer data was stolen. That distinction matters, even as the broader security implications remain troubling.
The Most Disturbing Part: The AI Agents Worked Independently
The investigation suggests the hacker did not personally perform every technical step.
Instead, the operator supplied relatively brief instructions, reportedly fewer than 100 Chinese characters, and allowed AI agents to investigate potential weaknesses.
The systems could examine websites, identify related infrastructure, test possible vulnerabilities, and pursue alternative approaches when initial attempts failed.
In one case described by the researcher, the agents reportedly moved beyond a company’s main application, located associated infrastructure, and eventually exploited a weakness that allowed malicious commands to be executed.
Successful intrusions typically required approximately 23 hours of automated activity.
The operation appeared to use a combination of DeepSeek and Anthropic’s Claude Code, according to Fagerland’s assessment.
Neither the use of those AI products nor their involvement in the reported campaign establishes that their developers authorized or intended the criminal activity.
But the broader development is difficult to ignore.
A cybercriminal can now potentially delegate much of the investigative and technical work to software that continues operating without constant human direction.
That changes the economics of cybercrime.
And economics can be just as dangerous as technology.
When a Successful Cyberattack Costs Around $50
Fagerland estimated the campaign consumed approximately $5,000 in AI processing tokens, equivalent to roughly $50 per successful intrusion.
Those are estimates, not independently audited financial records.
Nevertheless, consider the implications.
When attacking a website becomes cheaper, more automated, and less dependent on specialized expertise, criminals may be able to target far more organizations than before.
A small online retailer that previously attracted little attention could become one of thousands of automated targets.
A local business with outdated software could face repeated vulnerability testing without realizing anyone is looking.
And an organization might find itself confronting not one human attacker working through a list of targets, but software capable of investigating multiple systems over extended periods.
The threat is not that every AI-generated attack will succeed.
It is that even a relatively low success rate can become dangerous when the cost of attempting thousands of attacks falls dramatically.
This Is Bigger Than One Hacker
The Reuters investigation also describes other malicious activity uncovered by Fagerland, including a separate case involving stolen credit card information and an attempt to use AI agents against a cryptocurrency exchange.
Other cybersecurity investigations suggest the problem extends beyond this particular campaign.
Reuters reported that Gambit Security had documented AI-assisted intrusions involving multiple organizations and, in a separate September case, the theft of at least 600,000 payment-card records.
These were distinct incidents, not additional confirmed losses from the Chinese-speaking hacker’s campaign.
The pattern, however, raises an uncomfortable possibility.
The criminal underground is learning to use autonomous systems as operational tools rather than merely as assistants for writing code or generating phishing messages.
That development could make cybercrime more accessible to less-experienced attackers while allowing sophisticated criminals to expand their operations.
It could also overwhelm organizations that depend on limited cybersecurity staff and reactive defenses.
And that is where the conversation should become much more serious.
Who Is Responsible When AI Becomes the Weapon?
There is a difficult question that technology companies, lawmakers, and cybersecurity professionals will increasingly have to confront.
When a human criminal uses autonomous AI to attack hundreds of organizations, where does responsibility begin and end?
The criminal operator remains responsible for their actions.
But what obligations should fall on the developers of powerful autonomous tools? How should providers respond when their technology is misused? And can existing security practices keep pace with software capable of conducting attacks around the clock?
The developer of Cairn has publicly maintained that the platform was not created for malicious hacking.
That distinction is significant. Tools capable of finding security weaknesses can also help legitimate researchers protect computer systems.
The same general capabilities can serve defensive or criminal purposes depending on authorization and intent.
Banning every powerful cybersecurity tool would not eliminate the underlying problem.
But pretending that increasingly autonomous software creates no new risks would be equally irresponsible.
Canada Is Not Watching From the Sidelines
Canadian websites were among the targets identified in the reported campaign.
That should concern Canadian businesses, particularly smaller operators that may lack dedicated cybersecurity teams.
An online store, community organization, or local service provider can become vulnerable when outdated software, poorly protected administrative systems, or neglected third-party applications provide an entry point.
AI does not need to invent an entirely new vulnerability to cause damage.
It may only need to discover an existing weakness faster and more consistently than a human attacker.
Organizations should be reviewing software updates, administrative access, multifactor authentication, backup procedures, monitoring systems, and the security of outside vendors.
Consumers should also pay attention to account security, unique passwords, multifactor authentication, and suspicious financial activity.
These precautions cannot guarantee protection. They can, however, reduce exposure to many common attack methods.
The Real AI Revolution May Be Happening Where Nobody Is Looking
The public conversation about artificial intelligence remains dominated by productivity, employment, competition, and the promise of technological progress.
Those debates matter.
But the Reuters investigation points toward another reality: the tools being developed to automate useful work can also automate malicious activity.
A single operator allegedly directed a campaign against more than 1,200 websites while AI agents handled much of the technical process.
The reported expense was roughly $5,000.
That is not proof that every criminal can now compromise any website on demand. Nor does it mean cybersecurity defenses have suddenly become useless.
It does demonstrate why the old assumptions about the resources required to conduct sustained cyberattacks deserve reconsideration.
The next generation of cybercrime may not require larger criminal organizations. It may require fewer people with more capable machines.
And while governments debate how to regulate artificial intelligence, criminals are already experimenting with what these systems can do.
The uncomfortable truth is that technological capability often advances faster than the rules designed to contain its misuse.
The question is whether cybersecurity can adapt quickly enough.
Because the next major cyberattack might not begin with a room full of hackers.
It might begin with one person, a modest budget, and an instruction to an AI agent.
And by the time anyone notices, the software may have already moved on to its next target.
Chris Wick News — Independent news. Uncomfortable questions. No political team gets a free pass.
Sources
Reuters — Raphael Satter, October 19, 2026 How One Hacker Used AI Agents and a Few Thousand Dollars to Break Into Scores of Sites https://www.reuters.com/legal/government/how-one-hacker-used-ai-agents-few-thousand-dollars-break-into-scores-sites-2026-10-19/
Reuters — October 13, 2026 Anthropic Warns AI Models May Soon Self-Replicate and Could Wreak Havoc on the Web https://www.reuters.com/business/ai-models-may-soon-self-replicate-without-access-data-centers-anthropic-warns-2026-10-13/
