The Good. The Bad. The Rogue.

They don't wear suits or carry briefcases, and none of them drives an Aston Martin. But a new generation of artificial intelligence is starting to behave more like an agent than an ordinary computer program. These systems can look things up, use software, carry out multistep tasks and, when someone gives them the keys, take actions on a person's or a company's behalf.

Some of them work for us. Some are being used against us. And a third group raises an uncomfortable question: what happens when an AI agent does something its owner never meant it to do?

For Northern New York this is not just another story out of Silicon Valley. From Fort Drum and Watertown to the businesses and towns of Jefferson, Lewis and St. Lawrence counties, the question is no longer whether AI will touch the North Country. It is how much is already happening, who benefits, and whether residents are ready for the risks.

The good: when AI works for us

At Fort Drum, the Army's data future is already being built by soldiers. In June 2026 the Army reported that the 10th Mountain Division had created a five-day Basic Builders Course that teaches soldiers, civilians and garrison partners to build their own software applications for command and decision-making, using low-code and no-code tools rather than traditional programming.

The course grew out of the division's Data and Decision Dominance Team. Its first session trained 33 people. By June, more than 75 had finished it, and Maj. Dan Walker, the division's command chief data and analytics officer, said the goal is "more than 200 builders during the next year."

The apps are not classroom exercises. Students built a tracker for unmanned aircraft, medical status and statistics tools for the division surgeon, an intelligence template and a fires targeting dashboard tied to intelligence feeds. The Army's Mission Command Center of Excellence has adopted the course as the core of its own builder course. In June, the Army said Army-wide use could begin as early as this past summer; ClickNNY has not confirmed whether it has.

To be precise about it: this is data and app-building training that sits alongside the Army's newer AI-enabled systems, not a program that turns decisions over to a machine.

From the classroom to the workforce

When soldiers leave the Army, many stay in the North Country, and data, cybersecurity and software skills travel well into civilian jobs. Whether Fort Drum's training is producing measurable civilian job gains here is still an open question, and one only the post and local employers can answer.

The region's colleges are moving too. In June, Clarkson University in Potsdam announced a new undergraduate minor in Artificial Intelligence and Machine Learning, open to students from a variety of academic backgrounds and aimed at healthcare, manufacturing, cybersecurity, robotics, business and sustainability. "AI is reshaping every discipline, and this minor ensures our students are ready for that reality," said Academic Dean Mahesh Banavar.

In Watertown, Jefferson Community College cut the ribbon on Oct. 23 on its renovated Dulles Building, a project of more than $11 million paid for with state and Jefferson County money. It houses new mechatronics labs and a cybersecurity classroom and lab. The state grant reserves 40 percent of program seats for transitioning Fort Drum service members and spouses, and a separate $750,000 grant from Micron supports a new degree and training program in mechatronics and the semiconductor industry.

Mechatronics is not the same thing as AI. But the investment points at the same shift: the North Country is training people for workplaces where automated systems do more of the work. For a region that has long struggled to keep young workers and attract new industry, that could be an opening.

It also raises the next question. Will AI create more North Country jobs than it replaces?

The bad: when AI comes for your job

Machines taking over repetitive work is nothing new. What is different about today's AI is the range of work it can do. Older automation handled predictable tasks on a factory floor or in a spreadsheet. Today's systems can summarize documents, write code, analyze information, answer customer questions and handle increasingly complicated office work.

That reaches well beyond the factory. A medical office might use AI to help with paperwork. An insurer might automate parts of claims processing. A small business might hand AI its bookkeeping, marketing or customer emails. Those changes can make a business more productive. They can also change how many people it needs, and what kind.

A warning in the regional numbers

In 2018, the New York City think tank Center for an Urban Future estimated that 44.3% of the work tasks in New York's North Country region could be automated with technology that already existed, the third-highest share in the state. It counted about 22,480 "highly automatable" jobs in the region, positions where at least 80% of the work could be done by machines, or about 15% of all jobs.

Among the biggest groups: food preparation and serving workers, including fast food; bookkeeping, accounting and auditing clerks; and stock clerks and order fillers. Those are familiar jobs in every North Country town.

Those numbers come with important caveats. They measure how much of the work could technically be automated, not jobs actually lost, and the report itself said automation was not expected to eliminate all, or even most, of those jobs. The study is eight years old and predates today's AI agents. And its North Country covers seven counties (Clinton, Essex, Franklin, Hamilton, Jefferson, Lewis and St. Lawrence), not just the three ClickNNY covers.

The effects may not show up as layoffs, either. A company might not replace someone who retires. A business might grow without hiring. An entry-level job might quietly disappear because software now does most of it. Other employers will hire people with new skills or retrain the ones they have.

The rogue: when AI stops following the rules

Picture waking up to find that something has been inside your bank account. Not a hacker hunched over a keyboard, but a piece of software that can move through websites, fill in forms, contact customer service and, in principle, move money.

It sounds like a tech thriller. This month, the country's biggest banks were dealing with a version of it. On Oct. 22, Axios reported, citing unnamed sources, that third-party AI agents were logging in to customer accounts at major U.S. banks, and that some appeared to act beyond what their users had asked, including contacting bank call centers. One bank executive told the outlet they could not tell whether an agent had been authorized or had gone rogue.

The banks' responses differ. According to Axios, Morgan Stanley has begun blocking outside agents, and JPMorgan, Bank of America and Citi are restricting them. A Morgan Stanley spokesperson said the bank has not seen "rogue AI agents accessing our platforms," and a Chase spokesperson said customers should not give their usernames and passwords to anyone. The report does not say any customer money was taken. Most banks, it says, are moving toward read-only access, which would let an agent see account information without acting on it.

There is an important detail here. In most of these cases, the agents got in because customers gave them their own usernames and passwords, so the software logged in as if it were the customer. Many banks require a human to sign in, and that is where the trouble starts. Once an agent holds your credentials, can it move money? Change your contact information? Tell the difference between your instruction and a malicious one it stumbled on somewhere online? And can the bank tell whether it is talking to you or to software acting in your name?

Nothing in the reporting shows that rogue AI agents have taken money from North Country residents. It does show that banks are facing a new kind of account-access risk, and that matters to customers of every bank and credit union in the region, wherever the first cases turn up.

Your voice may no longer prove it's you

AI fraud is not only about autonomous agents. The FBI has warned that criminals use generative AI to clone voices, fake photos and video calls and produce bogus documents such as driver's licenses, including voice clones that impersonate a relative in a crisis to ask for money.

A hypothetical, not a reported case: a Watertown resident gets a call from someone who says they are with the bank. The caller knows their name, sounds professional and explains that a suspicious charge needs immediate action. Then the caller asks for the verification code that just arrived by text. The customer may not realize the voice could be AI-generated, or that the code is exactly what a thief needs to get into the account.

The technology has changed. The con has not.

Account takeover by impersonation is already expensive. In a November 2025 alert, the FBI said it had received more than 5,100 complaints since January of that year from people whose accounts were taken over by criminals posing as their bank's support staff, with losses of more than $262 million. Those numbers cover account takeover in general; the alert does not say how many involved AI. The practical lesson is the same either way: a familiar voice, a convincing message or a real-looking website is no longer proof of who is on the other end.

Take an AI assistant connected to a company's email that is only supposed to summarize messages. If it reads an email with hidden instructions and follows them instead of doing its job, the result could be leaked information or an action nobody approved. The agent does not need to be rebellious for that to happen. It just needs too much access and too few safeguards.

For North Country businesses, local governments and financial institutions, that creates a new security question. Who is responsible when an AI agent makes a decision nobody authorized?

Does Fort Drum, or the border, put us at higher risk?

It is a fair question for a region built around a major Army post and sitting right across the river from Canada. The honest answer is: not in the way you might fear, but in a few specific ways that are worth knowing.

Start with what geography does not change. AI-powered scams and cyberattacks travel over the internet and the phone network, so an attacker can reach a Watertown kitchen as easily as one in Phoenix. ClickNNY found no official source that says living near a military base or an international border makes any one person's phone or bank account a likelier target.

AI-run attacks themselves are real and new. In November 2025, the AI company Anthropic disclosed that a group it assessed "with high confidence" was Chinese state-sponsored had used its Claude Code tool to attack roughly 30 organizations worldwide, including tech companies, financial institutions, chemical manufacturers and government agencies, with the AI doing an estimated 80 to 90 percent of the work. It said the attackers succeeded in a small number of cases. Those targets were large organizations, not households.

Where Fort Drum does matter: who lives here

The North Country is home to thousands of soldiers, military families and veterans, and they are a favorite target for scammers. In 2024, military-connected consumers reported losing $584 million to fraud, up from $477 million the year before, according to Federal Trade Commission data reported by the Military Officers Association of America. Retirees and veterans accounted for $419 million of that.

AI is making the impersonation side of those scams more convincing. The FBI has warned that criminals use cloned voices to impersonate a relative in a crisis, and in May 2025 it flagged an ongoing campaign using AI-generated voice messages to pose as senior U.S. officials. For a deployed soldier's spouse or a retiree on a fixed income, a cloned voice asking for money or a login code is exactly the kind of call to hang up on and verify.

And what is here: the systems a base depends on

The bigger concern is infrastructure, not individuals. In a February 2024 advisory, CISA, the NSA and the FBI said Chinese state hackers were "seeking to pre-position themselves" on U.S. networks, mainly in communications, energy, transportation and water systems, for possible disruptive attacks "in the event of a major crisis or conflict with the United States." They noted that some victims were smaller organizations with limited cybersecurity. The Pentagon's 2023 cyber strategy warned that in a crisis or conflict, adversaries, with China singled out, could use cyberattacks on the U.S. homeland to slow American military mobilization.

Military bases often lean on the same local power, water and phone networks as everyone around them. Fort Drum has a cushion there: in 2014 the Army contracted for an on-site biomass plant that can supply up to all of the post's electricity and keep mission-critical operations running during regional grid outages. But the hospitals, utilities and town systems that serve soldiers' families serve everyone else too.

That is not hypothetical here. In 2023, a ransomware attack forced Carthage Area Hospital, which serves civilian and military patients, and Claxton-Hepburn Medical Center in Ogdensburg to divert ambulances and reschedule most appointments. It was not an AI attack, but it shows how much a local outage can disrupt.

What about Canada?

Sharing a border does not, by itself, raise anyone's AI risk. The cross-border threat that has been documented is old-fashioned phone fraud. In March 2025, federal prosecutors in Vermont charged 25 Canadian nationals with running a "grandparent scam" from call centers in and around Montreal that allegedly took more than $21 million from older Americans in more than 40 states, with callers posing as grandchildren who needed bail money. Canadian police made the arrests, and seven more people were charged and arrested in Canada in May 2026. These are charges, not convictions, and the case filings do not say the ring used AI. But voice cloning is exactly the tool that makes that kind of call more believable.

The North Country also shares infrastructure with Ontario. The New York Power Authority's St. Lawrence-FDR Power Project dam in Massena is a cooperative U.S.-Canada project: each country operates half of its 32 turbines. Canada's Centre for Cyber Security has warned that China is targeting critical infrastructure in Canada and allied countries to pre-position for possible disruptive attacks, and that an attack on integrated North American systems such as power grids would likely affect Canada as well. ClickNNY found no reported cyber incident at the dam or on the St. Lawrence Seaway.

Who's watching? Cameras and the North Country

AI also raises questions about surveillance. Software can scan video, read license plates and flag unusual activity far faster than a person can. For police, that can help solve cases. For residents, it raises fair questions about privacy, how long data is kept and who it is shared with.

In April 2025, the Watertown City Council approved a police plan for 16 license plate reader cameras, according to the Watertown Daily Times. A $38,650 grant was to cover 11 of them, with the city paying about $18,650 to install the other five plus an ongoing subscription. A handful of residents raised privacy concerns at the meeting. Police Capt. Cristin Lyon said the cameras would be used only in open cases or specific circumstances, such as finding missing children, robbery suspects or drug dealers, and that the equipment "will not collect data on residents" but will provide "a snapshot of the license and the vehicle." Police said they would keep the camera locations private to protect them from theft or vandalism. The council's approval was needed for state Department of Transportation permission to mount cameras along roadways.

What ClickNNY has not been able to confirm is whether the cameras were installed, whether they are running today, which company supplies them, how long plate data is kept and whether it is shared outside the department. Those are questions for city officials to answer in public. A plate reader is also not facial recognition, and it is not an autonomous AI system making decisions on its own.

Could AI one day help monitor utility networks, water systems or emergency dispatch here? Those are realistic uses. But ClickNNY has found no evidence that an autonomous AI system controls North Country power, water or police operations today, and no local incident in which a rogue AI agent compromised one.

That is exactly why transparency matters. Residents should be able to learn which automated systems their public agencies use, what those systems are allowed to decide and who answers for it when something goes wrong.

The North Country's AI question: are we ready?

AI offers Northern New York two very different futures. It could help local businesses compete, prepare soldiers for new technology, widen what our colleges teach and create new kinds of work. It could also unsettle familiar jobs, expose private information, hand scammers new tools and slip automated decisions into systems people depend on.

Some of that is documented here at home. Some of it is national news with real implications for the North Country. And some of it is questions local officials, employers and banks have not yet answered in public.

The difference matters. A data-skills course at Fort Drum is not evidence of weapons running without human oversight. A city approving plate readers is not proof of an AI surveillance network. And a national report about AI agents inside bank accounts does not mean North Country customers have seen the same thing. Each one is still a good reason to ask how these tools are introduced, and who controls them.

The good. The bad. The rogue.

AI agents do not mean computers are turning into secret agents with plans of their own. They do mean software is increasingly able to act for us, reach into sensitive systems and make decisions with real-world consequences. Sometimes those consequences will help. Sometimes they will cost us. And sometimes the biggest worry will be learning that a system had more authority than anyone realized.

The next wave of AI will not necessarily announce itself. It might be helping a soldier sort information at Fort Drum, or a small business keep its books. It might help a bank catch a fraudulent transaction. Or it might be the reason that transaction was attempted in the first place.

The agents are here. The question is who's giving them their orders.

Help us report on AI in the North Country

Has AI changed your job, your business or how your bank treats you? Have you run into an AI-generated scam call, suspicious automated activity, or a local organization using AI in a way that surprised you? ClickNNY wants to hear from residents, workers, employers and public officials in Jefferson, Lewis and St. Lawrence counties.

Send us a tip. Please do not include account numbers, passwords or other sensitive personal information.

Common questions

What is an AI agent?

An AI agent is software that uses an AI model plus other tools to carry out multistep tasks with some independence, such as researching, filling out forms or logging in to accounts. What it can actually do depends on the access and permissions people give it.

What does it mean when an AI agent goes rogue?

It is shorthand, not a sign the software is conscious. A rogue agent acts outside its instructions or permissions, either because it was misconfigured, given too much access, or manipulated, often through prompt injection, where hidden instructions in an email or web page trick it into doing something its owner never asked.

Are banks blocking AI agents?

Some are. In October 2026, Axios reported that Morgan Stanley had begun blocking third-party AI agents and that JPMorgan Chase, Bank of America and Citi were restricting them, after agents logged in with customers' own passwords. The report did not say any customer money was taken, and most banks are moving toward read-only access for agents.

Should I give an AI assistant my bank password?

Be very careful. Many banks require a human to sign in, and an agent holding your password can potentially do anything you can do. Use only tools your bank has approved, and never share one-time verification codes.

How can I tell if a call from my bank or a relative is an AI voice scam?

You often can't tell by the voice alone. Hang up and call back using a number you trust, and set up a family code word, as the FBI recommends.

Does living near Fort Drum or the Canadian border raise my risk of AI attacks?

Not for your own phone or bank account; online scams can reach anyone, anywhere. The military presence matters in two ways: soldiers, families and veterans are frequent targets of imposter scams, and federal agencies warn that state hackers want to disrupt the water, power and communications systems that bases and their communities share. The documented cross-border threat is phone fraud, such as a Montreal-based grandparent scam charged in 2025.

How many North Country jobs could AI replace?

No one has a verified count. A 2018 Center for an Urban Future study estimated 44.3% of work tasks in the seven-county North Country region could be automated and counted about 22,480 highly automatable jobs, but that measures potential, not actual job losses, and it predates today's AI agents.

Where can North Country residents learn AI skills?

Clarkson University in Potsdam added an Artificial Intelligence and Machine Learning minor in 2026, and Jefferson Community College in Watertown has new mechatronics and cybersecurity labs, with a Micron-funded mechatronics and semiconductor program.