The Security Story Governments Actually Bought
Washington dismissed the industry's "fear the models" pitch, and on both sides of the Atlantic "protect the machines" became the stated reason for rules that override local voters and the labs' own safeguards.
Over two years, Britain's government sat down with 137 businesses about data-center policy, and with two community groups and no environmental organizations at all [1]. What came out of that room was a classification: data centers are now "nationally significant infrastructure," which means a project can skip local planning and go straight to a minister for approval — the same government whose finance minister calls its data-centre covenant "built on sovereignty and security" [2]. The growth-zone designation, an earlier instrument in the same family, shows how the work is done. At Lanarkshire it pushed an £8.2 billion CoreWeave-DataVita complex over green-belt objection [3], while the government privately knew the site's power story was not the one it was telling.
AI is critical to the UK’s future prosperity and security – and to unlock its benefits, we need the infrastructure that underpins it. — Government of the United Kingdom
This was the endpoint of a fight between two security stories. September 2026 was the contest's endpoint, not its start: when OpenAI and Anthropic formally asked Washington to regulate the existential risk from their own models, the White House answered "HOAX" and resisted regulation [4]. The winning story had been selling in plainer terms for more than a year before that answer. In October 2025 OpenAI asked the White House to back its $500 billion Stargate expansion by casting a widening "electron gap" with China as a national-security threat [5] — the same filing that projected a trillion dollars of AI infrastructure lifting GDP growth more than 5 percent, even as new data centers pushed household electricity bills up in 13 states. Nvidia's Jensen Huang repositioned his company as infrastructure, not a technology firm.
not a technology company only anymore — Jensen Huang
That frame fed CISA's security guidance into momentum on Capitol Hill to formally designate AI infrastructure as critical [6]. AWS chief Matt Garman went further, recasting the opposition itself as the work of foreign adversaries.
this buildout is so important geopolitically that there are widespread reports of various countries intentionally seeding misinformation in the U.S. about data centers to trick us into slowing down. — Matt Garman
By his tally, anti-data-center movements had blocked $68 billion in projects in one quarter, with 100 moratoriums under consideration nationwide [7]. The people he is describing are, by one poll, 74 percent of Wisconsin voters who say the machines cost more than they give back [8]. The instruments carry the story as their stated reason. The Energy Department put four federal sites — Idaho National Laboratory, Oak Ridge, Savannah River and Paducah — forward for AI data centers with streamlined environmental review, and its secretary named the project's rationale for national security.
By leveraging DOE land assets for the deployment of AI and energy infrastructure, we are taking a bold step to accelerate the next Manhattan Project—ensuring U.S. AI and energy leadership. — Alfie Moon
The EPA, in the same stretch, exempted islanded power plants serving a single site like a data center from Clean Air Act rules and proposed dropping public notice-and-comment for minor sources — even as Maryland counties enacted bans and pauses, Pennsylvania's governor imposed a moratorium, and Texas's governor blocked new data centers from the state grid [9]. The same frame points inward, and there it cuts the labs' own rules. A U.S. export-control directive forced Anthropic to disable two of its models for every foreign-national customer [10].
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. — Anthropic
The disruption reached Tata Consultancy Services, Infosys and Indian startups, and pushed India to accelerate its own sovereign models [10]. Months earlier, the inward edge had cut harder. The Pentagon threatened to designate Anthropic a "supply chain risk" — the label used to exclude vendors — unless the lab removed safeguards blocking mass surveillance and fully autonomous weapons [11]. Anthropic held out, and the classified work went to Elon Musk's xAI to run Grok, while Google and OpenAI negotiated under the same "all lawful purposes" standard [12]. Against all this, the record holds two stops, and both were adjudicative rather than political. A federal appeals board halted the $3 billion Townsite data center near Boulder City, Nevada, finding the Bureau of Land Management acted "arbitrarily or capriciously" when it recycled a 2023 solar-project environmental review to approve a 167-megawatt data center on federal land [13]. And U.S. District Judge Rita Lin ruled that the government's exclusion of Anthropic was unlawful retaliation violating the company's First and Fifth Amendment rights [12]. The political channel, by contrast, recorded deadlock and veto rather than stops. Little Rock's board split 4-4 on a moratorium and cleared a $1 billion Google facility [8]. Pulaski County's judge vetoed an 11-4 permitting ordinance, with a $6 billion AVAIO Digital campus pending [14].
The County should not create obligations today and determine tomorrow whether it has the people and money to fulfill them — Barry Hyde
What the two rulings share matters more than what they stopped. Both came from inside the state's own rulebooks — a board applying review standards, a judge applying constitutional rights — and both turned on how the toolkit was wielded, not on whether it should exist. Neither touched the security story itself, which still carries every instrument on the record.
- 1. UK Classifies Data Centers as Nationally Significant Infrastructure
- 2. John Healey Launches British Data Centre Covenant for AI
- 3. UK Government Designates Lanarkshire AI Growth Zone Amid Energy Disputes
- 4. OpenAI and Anthropic Call for AI Existential Risk Regulation
- 5. OpenAI Urges Federal Support for $500 Billion Stargate Project
- 6. CISA Urges Secure Infrastructure for AI Data Centers
- 7. AWS CEO Warns Data Center Opposition Threatens US AI Lead
- 8. US Cities and States Clash Over Data Center Moratoriums
- 9. EPA Eases Rules as States Block AI Data Centers
- 10. India Pursues Sovereign AI After US Bans Anthropic Models
- 11. Pentagon Deploys xAI Grok After Standoff With Anthropic
- 12. Defense Department Launches ChatGPT Mil and Grok for Government
- 13. Judge Halts Nevada Data Center Over Environmental Review Failures
- 14. Pulaski County Judge Vetoes Data Center Regulations