AI Firms Targeted as Data-Center Gear Faces Theft

A cybersecurity startup founded by former Department of Government Efficiency staffers told Anthropic that its researchers had penetrated sensitive internal data and messaging systems at the company behind the Claude chatbot. The startup, Cathedral, said the work took 12 hours, according to a draft report reviewed by The Washington Post. That’s a corporate security claim with a sharp power imbalance built in: researchers reached systems holding sensitive information, while the company responsible for those systems now faces the consequences.

Inside the Companies’ Defenses

The report was published October 26, 2026, by Elizabeth Dwoskin and Gerrit De Vynck. The account identifies Cathedral as a months-old startup and says it told Anthropic its researchers had accessed internal data and messages. It doesn’t describe what information they saw, how the access happened, or how Anthropic responded. Those limits matter: the reported claim is serious, but the details available here stop at what Cathedral told the company and what the draft report described.

A separate incident put a direct bill in front of startup founder Colin Parsons. AI-powered hackers targeted him at 11 p.m. on a recent Friday night. By the next morning, he had about 50 emails saying he’d been charged a few more dollars for services from Anthropic and OpenAI. The charges totaled $449. The account gives a small, concrete measure of the cost when digital systems become a route for others to spend: Parsons had the messages and the bill, while the companies named in them supplied the services being charged.

These are separate cases in the account, not evidence of one coordinated operation. But both place AI companies and the people using their services inside a contest over access, data, and money. The available reporting doesn’t identify a grassroots response, mutual aid effort, or horizontal organizing around either incident. No such response is described here.

The Hardware Has a Price Tag

Organized criminal groups are increasingly targeting valuable data-center equipment through deceptive techniques, NBC News reported. In early June, a trailer carrying about $13 million worth of computer server equipment was headed east toward Fayetteville, Georgia, home to one of the country’s largest data-center developments. Trucking, supply-chain, and security industry veterans described dramatic robberies connected to the growing theft of technology equipment.

NBC News reporter Lisa Cavazuti wrote the report. The account links the thefts to equipment with a stated value of millions, but it doesn’t provide details about who owned the trailer, who was targeted in particular robberies, or what happened to the equipment headed toward Fayetteville. What it does show is the physical side of the AI business: server gear moves through supply chains, and criminal groups use deception to target it.

Washington’s Proposed Fix

Tal Feldman and Ibrahim Dagher argued in a Washington Post opinion article that the AI race has a cybersecurity problem and that Washington has tools to push AI companies to take threats seriously. Feldman is a Yale Law School student who previously built artificial intelligence models for the U.S. government. Dagher is pursuing law and PhD degrees at Yale and writes about AI policy and national security.

Their argument places a proposed answer in Washington’s hands: pressure from the state, directed at companies. The account doesn’t spell out which tools the writers mean or report that any have been used. It also doesn’t describe elections, legislation, or community-led protections. The people and institutions named here are companies, researchers, a founder, industry veterans, and policy writers; the proposed lever is governmental pressure on AI firms. The companies’ systems, the founder’s $449 bill, and the multimillion-dollar shipment remain the hard particulars.