Nick Lawler, general manager of the public utility behind the power lines and water system in Littleton, Mass., hung up on the FBI agent who called in November 2023 to tell him Chinese government hackers were inside the utility's network. The agent called Lawler on a Friday at the start of Thanksgiving week and asked for Lawler's personal email address so the agent could send him a link. "It sounded like one of those Microsoft $MSFT scams," Lawler told The Register. He swore at the agent, refused to click the link, and hung up. Then he called the FBI's Boston office himself, and the same agent answered.
The hackers were Volt Typhoon, a group backed by the Chinese government. Lawler couldn't see why it would bother with a utility serving Littleton and neighboring Boxborough, two towns with about 15,000 people between them. "We don't own transmission. We're a distribution company. Yes, we're part of the overall grid, but the impact of taking out Littleton is small. You would never think that would be a target of any type of attack," he said.
The hackers got in through a flaw in a FortiGate firewall that its maker, Fortinet $FTNT, had fixed in December 2022, but the outside company Littleton paid to manage its computers never installed the fix. They had been inside for about 10 months by the time the FBI called, according to WaterISAC, a group that shares security information among water utilities. "I wouldn't say anything related to our substation or our engineering was compromised. They did access our servers," Lawler said. Sensors from the industrial security firm Dragos, which went onto the utility's equipment network in August 2023, had picked up traffic that didn't belong there, including communications with China, and once the FBI called, their logs showed when and where the hackers had moved. "That's when it became real for us," he told the American Public Power Association, which ran the government-funded program that paid for the sensors.
Littleton's break-in is the kind of attack that people picture when they warn AI could wipe out humanity, only with a machine doing the hacking. An AI can already do much of that work: Anthropic said in November 2025 that a group it believes the Chinese government backs used its Claude Code tool to perform 80% to 90% of a hacking campaign. That AI-run campaign went after about 30 organizations, according to the SANS Institute, a cybersecurity training organization. But an AI trying to black out a town would run into the same defenses grid operators have spent years building against human hackers like Littleton's, and those defenses are strongest at the country's biggest power companies and thinnest at small utilities like Littleton's.
The grid defenses built after Ukraine's 2015 blackouts
Many of those defenses guard against the mistakes hackers exploited in Ukraine on Dec. 23, 2015, when they took over the control screens at three power companies and cut power to about 225,000 customers. At about 3:30 p.m., a worker at one of the companies, Prykarpattyaoblenergo, watched his cursor move by itself and open the breakers at a substation, and when he grabbed his mouse, the system logged him out. The hackers had used stolen logins to come in through VPNs that linked each company's office network to its grid control systems. The VPNs appear to have asked for nothing more than a password, and no one was watching the control networks for intruders, according to a 2016 analysis by the electricity industry's information-sharing center and the SANS Institute.
The largest U.S. grid operators now have to wall off their most critical equipment and ask for more than a password to reach it. The North American Electric Reliability Corporation (NERC) writes the grid's security standards, and its rules require the computers running the grid's most critical equipment to sit inside what NERC calls an "electronic security perimeter" — a walled-off zone whose every connection passes through a gate, usually a firewall, that blocks traffic the utility hasn't approved. Anyone logging in from outside has to go through a separate computer, use an encrypted connection, and prove who they are with a second check beyond a password, exactly the check Ukraine's power companies missed.
But a firewall can't catch someone who is already inside it. In June 2025, federal regulators approved a rule that makes grid operators watch the traffic between their own control computers. The Federal Energy Regulatory Commission (FERC) said the old rules mostly watched the edge, letting an attacker who got past it take over administrator accounts and move around unnoticed. Volt Typhoon did exactly that, staying hidden by logging in with real employee accounts, using the computers' own built-in tools, and deleting the logs.
The biggest test of those defenses since Ukraine came in Poland on Dec. 29, 2025, and the attackers got in much the way Ukraine's had. Every site they targeted had a Fortinet FortiGate firewall open to the internet, still on its factory password and with no second login check, according to SecurityWeek. At wind and solar farms, the attackers wiped files the safety equipment needed and locked the farms' own staff out of their network devices, according to Ampyx Cyber, a U.S. security consultancy that summarized Poland's investigation. The turbines kept spinning, but the local grid's operators could no longer see or control them, and none of the farms had software watching their computers for intruders. At a large heat and power plant that did have it, the software caught and blocked the attackers' data-wiping program.
The small grid sites the strictest rules leave out
The strictest U.S. rules cover big power plants and high-voltage lines, and the wind farms, solar farms, and town utilities that hackers keep getting into mostly fall outside them. Nozomi Networks' field chief information security officer, Markus Mueller, thinks the U.S. has the world's best energy security rules but that they skip smaller sites. "If the site isn't big enough, it's not going to have requirements from a compliance standpoint," he told Latitude Media. One small site matters little to the grid, Mueller said, but hitting many at once, as in Poland, can still threaten it.
The June 2025 rule that makes grid operators watch the traffic between their own control computers covers only the systems NERC rates most important to the grid, plus middle-rated systems that can be reached from outside the company's network. A company with neither doesn't have to follow it at all. Littleton got its own monitoring sensors only through a government-funded program. "As a small utility, we don't have that type of staff to build that network," Lawler, the general manager, told the American Public Power Association.
Robert M. Lee, CEO of Dragos, the security firm whose sensors caught the hackers in Littleton, told The Record in February that several of the biggest U.S. electricity companies can now find and remove Volt Typhoon, the Chinese government-backed group that broke into Littleton's utility. But he doesn't expect water utilities and small public utilities ever to catch up. Volt Typhoon has already shown how far it can get at a small utility. At one unnamed water utility, it spent nine months working its way to a server that held saved logins for the utility's treatment plants, its wells, and a power substation. "In my assessment … we're going to have to live with the reality that a portion of our infrastructure is currently compromised and will remain compromised at the current trajectory of the community," Lee said.
An AI would make getting through those gaps faster and cheaper. In the campaign Anthropic described, Claude mapped its targets, wrote attack code, and stole logins with people stepping in only now and then. It also invented passwords that didn't exist, errors Anthropic said still stand in the way of attacks that run without people. "It definitely lowers the bar for the amount of operational knowledge that you need to make an attack," he said.
None of the hackers in Ukraine, Littleton or Poland needed a clever new flaw to get in. In Ukraine they came through connections that asked only for a stolen password, in Littleton through a firewall flaw its maker had already fixed, and in Poland through firewalls still on their factory passwords. An AI would need the same kind of opening.
Regulators have ordered the monitoring extended to the computers that control who can log in, but no one has reported that those rules have been filed. For the families who get their power and tap water from a town utility, catching an intruder falls to the few staff the utility can afford, and an AI would give those staff less time to notice than a human hacker would.