Oct 12, 2026

Drones Hit Yandex's AI Data Centers as US Restores UK Access to Frontier Models

Ukrainian drones knocked a third Yandex AI data center offline in four days, while the White House ended a 17-day freeze on UK pre-release testing of US frontier models.

Listen to this update

The physical and political plumbing of artificial intelligence dominated the weekend. Ukrainian drones knocked a third Yandex data center offline in four days, Washington quietly reversed a freeze that had kept Britain's AI testers away from the newest US models, and fresh money and fresh opposition kept piling up around the data centers that power the AI boom.

Yandex's Vladimir data center goes dark

Russian tech giant Yandex said on Sunday, October 11, that a drone attack had hit its data center in Vladimir, east of Moscow, forcing it offline. No injuries were reported, according to Al Jazeera. In an alert to customers, the company said data center operations were fully suspended, its cloud platform was running in emergency mode, and the resources still running were unstable. It advised clients to activate their disaster-recovery plans and warned that support channels could also be disrupted.

The strike was the third on Yandex facilities within a week. On October 8, an attack shut down the company's Sasovo hub in the Ryazan region, which Al Jazeera reports houses two of the three supercomputers Yandex uses to develop its AI models. Baxtel, which tracks data center infrastructure, described Sasovo as the first major Russian data center taken offline during the war, with Yandex Cloud reporting a power outage in one of its availability zones. A day later, a Yandex site in the Kaluga region was partly knocked out.

The attacks come as Russia has struck Ukrainian data centers and telecom sites over the past month, including targets in central Kyiv, causing temporary internet outages for roughly 100,000 households. Ukrainian President Volodymyr Zelenskyy said on October 8 that Kyiv answers such strikes in kind: "We always respond in mirror-like fashion." The campaign marks a clear shift: AI compute is now being treated as strategic infrastructure, and a military target.

US restores UK access to frontier models after 17-day freeze

The White House has dropped a request that kept the UK's AI Security Institute (AISI) from testing the newest American frontier models before release. Startup Fortune and Resultsense, citing reporting by The Telegraph, said the reversal came on October 11 and 12, ending a pause of about 17 days.

The freeze began on September 24, when the Office of the National Cyber Director asked OpenAI and Anthropic, and according to some reports Google DeepMind as well, to hold new models back from British testers until US agencies had reviewed them first. Anthropic complied, keeping its Claude Mythos 5.1 model away from AISI. UK AI minister Kanishka Narayan reportedly lobbied for the decision to be reversed.

The episode exposed how fragile voluntary testing arrangements are. AISI has no statutory right to pre-release access, and a single directive from Washington was enough to cut it off. That debate is gathering pace: in an ITV interview last week, Microsoft president Brad Smith called for mandatory incident reporting, independent pre-launch safety testing and government powers to order AI systems offline, saying, "I think the time to act is now."

SWI raises €800 million to turn a bitcoin miner into AI capacity

Amsterdam-listed SWI Group said it raised €800 million (about $894 million) from institutional investors at a €6.1 billion valuation, CoinDesk reported on October 12. The money will fund data center and AI computing projects in Europe and the US, including converting the US and Swedish sites of Genesis Digital Assets, a former FTX-backed bitcoin miner in which SWI gained voting control in August, to AI workloads. SWI plans to sell the unit's sites in Kazakhstan. CoinDesk notes that bitcoin miners have now signed more than $70 billion in AI and high-performance computing contracts, following similar moves by TeraWulf and Core Scientific.

The data center backlash keeps building

Not every community wants that capacity. TechCrunch's weekend Equity discussion looked at Amazon's recent pledge to stop using nondisclosure agreements when negotiating data center deals with local governments, following a similar move by Microsoft earlier this year. The hosts argued that secrecy has fueled public distrust, with ending NDAs a useful first step but unlikely to win over opponents on its own. The pushback is real: New York has a moratorium on permits for large projects while an environmental review is completed, and Fortune profiled bipartisan opposition to a $165 million, 1,300-acre Amazon-linked project in Hazle Township, Pennsylvania.

Why it matters

This weekend's stories point the same way. As AI gets more capable, the hardware behind it is being fought over: bombed in wartime, financed at multibillion-euro valuations, contested in town halls and controlled through national-security directives. Who gets to build compute, and who gets to test what runs on it, is quickly becoming as important as the models themselves.