By SUZANNE DOWNING
Sept. 11, 2026 — The artificial intelligence company whose employees have poured hundreds of thousands of dollars into Jonathan Kreiss-Tomkins’ campaign for Alaska governor now says its own technology was used by an Iran-linked operation to help target US Navy warships.
Anthropic disclosed that an Iran-linked account used its Claude artificial intelligence model to gather information about American naval forces in the Middle East and develop potential targeting recommendations.
According to The Wall Street Journal, the operation used Claude to compile targeting handbooks, track naval positions and search for vulnerabilities in shipboard systems.
The information included names of American military personnel gathered from public photographs, ship and aircraft transponder identifiers, commercial satellite-imagery searches and lists of websites that reveal naval movements.
The Iran-linked operator also used Claude to research vulnerabilities in maritime satellite terminals, Cisco communications equipment and industrial-control products, Anthropic reported. The company said it shut down the account but could not determine whether the intelligence was used in an actual Iranian military operation.
The disclosure was part of a much broader Anthropic threat-intelligence report documenting the use of Claude by hostile governments, weapons researchers, cybercriminals and surveillance operations.
A cell in northern Yemen used Claude to help develop guidance software for rockets and a planned ballistic missile with a range of more than 2,000 kilometers. The group also explored a missile design incorporating a hypersonic glide vehicle.
“Our safeguards blocked many of their requests, but not all of them,” Anthropic reported.
Other actors used Claude to develop technology for autonomous Russian drone swarms, research weapons for China, conduct cyberattacks and build surveillance systems targeting political dissidents, Uyghurs, Tibetans, religious leaders and Taiwanese political figures. One China-based actor reportedly used Claude to create a 200-page proposal for an anti-torpedo system intended for the Chinese navy.
The disclosures demonstrate that even a company that has carefully built its public image around artificial intelligence safety cannot keep its technology from being exploited by America’s adversaries.
That makes Anthropic’s extraordinary financial footprint in Alaska’s 2026 governor’s race more relevant than ever.
Six Anthropic employees contributed a combined $372,000 to Kreiss-Tomkins between February and May, according to campaign-finance reports. That included a $100,000 contribution from Anthropic employee Drake Thomas. Daniel Ziegler, Jan Leike, Steven Bills and Evan Hubinger were also among the candidate’s largest donors.
The Alaska Story previously identified at least $521,812 in contributions from people connected to the artificial intelligence industry, representing more than 28% of the approximately $1.85 million Kreiss-Tomkins had raised at that point earlier this summer.
A later San Francisco Standard analysis found an even wider network. In addition to the $372,000 from six Anthropic employees, Kreiss-Tomkins received $273,000 from people working at AI research and safety organizations and another $160,765 from employees of nonprofits and policy groups influential in the AI field.
Their size, concentration and common industry connections are unprecedented in an Alaska governor’s race.
The San Francisco Standard reported that some donors knew Kreiss-Tomkins through AI circles and wanted to help elect someone who shared their concerns. Kreiss-Tomkins signed an artificial intelligence safety letter from the Future of Life Institute nearly a decade ago.
That network now deserves greater scrutiny because Kreiss-Tomkins is proposing an immediate statewide moratorium on new data centers until the Alaska Legislature adopts a regulatory framework governing them.
The policy raises several questions that have not been answered.
Who helped Kreiss-Tomkins develop the proposal? Has he consulted Anthropic employees, executives or affiliated AI policy organizations? Have any of his large donors advised his campaign on data centers, artificial intelligence regulation or liability rules? And who would benefit from preventing new computing infrastructure from being built in Alaska? Anthropic?
A moratorium could block new competitors and independent developers from establishing an Alaska presence while leaving the largest existing artificial intelligence companies — including Anthropic and its powerful cloud-computing partners — in control of their established infrastructure elsewhere.
It could also delay projects designed to use stranded North Slope natural gas, including a proposed $500 million AI and cloud-computing center that would generate its own power and operate independently of the Railbelt electrical grid.
The US Air Force has separately invited proposals for as many as 12 artificial intelligence data centers on military property at Joint Base Elmendorf-Richardson, Eielson Air Force Base and Clear Space Force Station. It remains unclear whether Kreiss-Tomkins’ moratorium would reach projects on federal military installations.
JKT’s proposal would not simply regulate water consumption, protect residential ratepayers or require dedicated power generation. It would begin with a blanket prohibition applying to every new data center, regardless of its location, power source, size or connection to a public utility.
A moratorium may appeal to Anthropic employees who believe the expansion of computing capacity should be slowed. Whether it would benefit Anthropic as a company is less clear, but it could protect the competitive position of established AI giants by making it more difficult for new projects and potential rivals to obtain computing infrastructure.
The internal alarm at Anthropic became public this week when researcher Jacob Coxon resigned after working for three years in pretraining research at Anthropic and OpenAI.
“Neither company is acting responsibly,” Coxon wrote. “They are racing straight to self-improving superintelligence and gambling with our lives.”
Coxon said employees building advanced AI privately fear that the technology could escape human control, hack critical systems and acquire power and resources. He described Anthropic as being locked in a race to reach superintelligence first because company leaders do not trust competitors to handle the technology responsibly.
He called that competition a “hubristic gamble” and suggested that preventing a global race may require a temporary ban on further increases in AI model capabilities.
One thought on “Anthropic says Iran used Claude to target US warships as company’s money network shapes Alaska governor’s race”
My question is why anyone from the Lower 48 would be so concerned about Alaska’s race for governor that they would individually contribute what amounts to a yearly salary for most of us to a bizarre candidate. I assume these were coordinated contributions and backed by their companies, rather than their personal expenditures. And here the “no on 2” folks say we don’t have dark money in our campaigns. /spit