Public Policy & Governance
Top Line
New York's state legislature has passed a one-year moratorium on hyperscale datacenters exceeding 20MW, with the bill now awaiting Governor Hochul's signature — the first such state-level action in the US and a direct legislative constraint on AI infrastructure expansion.
Seattle's city council voted unanimously to enact a year-long ban on new datacenter construction, making it the largest US city to impose such a moratorium and signalling that municipal-level AI infrastructure regulation is now an active policy tool, not just rhetoric.
The UK's Ministry of Justice will trial AI legal assistants in crown courts, announced by Deputy PM David Lammy — a concrete public sector adoption move that is already drawing opposition from the legal profession over funding substitution concerns.
The US Senate Judiciary Committee is scheduled to vote on the NO FAKES Act, which the EFF argues would create broad internet censorship mechanisms rather than targeted privacy protections, placing it in direct tension with existing Section 230 frameworks.
The UK Medical Protection Society has formally warned ministers that current medical negligence law creates liability exposure for doctors and the NHS over AI diagnostic errors, calling for statutory reform — a gap between AI adoption policy and legal infrastructure that applies across multiple jurisdictions.
Key Developments
US Sub-National Governments Move to Restrict AI Datacenter Expansion
Two distinct legislative actions this week mark a significant shift in how sub-national US governments are responding to AI infrastructure pressure. New York's state legislature passed a one-year moratorium on hyperscale datacenters over 20MW, driven by State Senator Kristen Gonzalez, and the bill now sits with Governor Kathy Hochul for signature or veto. Separately, Seattle's city council enacted its own year-long ban by unanimous vote. These are enacted legislative measures — not proposals or consultations — though the New York bill remains contingent on the governor's action. The Guardian The Guardian
The policy rationale in both cases centres on energy and water resource stress, with a Guardian analysis finding approximately two-thirds of planned US datacenters are sited in drought-affected areas. The Guardian This is the first time environmental resource constraints — rather than safety, privacy, or competition concerns — are driving hard AI-related legislative action at scale in the US. The moratoriums are explicitly temporary, framed as pause-and-plan measures rather than outright prohibition, but they set a precedent that other state and municipal legislatures are likely to reference. The federal government has no equivalent mechanism in force, creating a patchwork regulatory environment that will complicate hyperscaler investment planning.
UK Government Advances Public Sector AI Adoption in Courts, But Legal Framework Lags
Deputy Prime Minister David Lammy is set to announce a trial of AI legal assistants in crown courts in England and Wales, framed as a response to the chronic case backlog. This is a confirmed government announcement of a live pilot — not a consultation or a strategy paper. The Bar Council and legal professional bodies have responded with immediate warnings that the technology must not be used as a substitute for court funding and staffing, a concern that reflects a pattern seen across public sector AI deployments where technology becomes a cost-reduction alibi. The Guardian
Running in parallel, the Medical Protection Society has formally warned ministers that English negligence law has not been updated to account for AI-generated clinical errors, meaning liability currently flows to clinicians and NHS trusts regardless of whether the AI system caused the harm. The Guardian This is a structural implementation gap: the UK government is actively deploying AI in public services while the legal liability framework governing those services has not been reformed. This gap is not unique to the UK — comparable issues exist in EU member states operating under the AI Act's high-risk system provisions — but the UK's post-Brexit regulatory independence means there is no supranational backstop forcing resolution.
NO FAKES Act Advances in Senate Judiciary — Scope and Censorship Risks Contested
The Senate Judiciary Committee is scheduled to mark up and vote on the NO FAKES Act (Nurture Originals, Foster Art, and Keep Entertainment Safe Act), which would create federal rights over AI-generated replicas of individuals. The Electronic Frontier Foundation has issued a formal opposition brief arguing the bill's current drafting extends well beyond targeted non-consensual intimate image harms into a broad platform liability regime that would incentivise over-removal of lawful content and layer censorship mechanisms on top of existing legal and voluntary takedown frameworks. EFF
The bill's advancement comes in the same news cycle as a UK Labour MP's lawsuit against xAI over Grok-generated sexualised images The Guardian and Prime Minister Starmer's public endorsement of that legal action The Guardian, illustrating transatlantic convergence on the political salience of synthetic media harms even where legislative approaches diverge. The EFF's core objection — that NO FAKES conflates privacy harms with intellectual property protection in ways that damage free expression — is a substantive drafting critique, not opposition to the underlying harm. This distinction matters for policy professionals assessing whether the bill is refinable or fundamentally flawed.
AI Profit Distribution Enters Legislative Debate in the United States
Senator Bernie Sanders' proposal for a sovereign wealth fund capturing a share of AI company equity has moved from op-ed into active congressional debate, with Politico reporting that the broader fight over AI profit redistribution is gaining traction across party lines even as a specific White House meeting on the topic appears uncertain. Politico This is a political agenda item, not enacted legislation, but it represents a shift in the Overton window — taxation and public ownership of AI returns are now legitimate subjects of congressional deliberation rather than fringe positions.
The debate has an international dimension: Australian Senator David Pocock is publicly advocating for a 25% levy on AI datacenter operators using Australian land, energy, and water, explicitly invoking the precedent of under-taxed gas exports. The Guardian These are opinion and advocacy positions rather than enacted measures, but the simultaneous emergence of fiscal redistribution arguments in the US and Australia — both framed around natural resource analogies — suggests a coordinating policy narrative is developing that legislative actors in multiple jurisdictions are beginning to test.
Signals & Trends
Environmental and Resource Regulation Is Becoming the Fastest-Moving Track for AI Governance
While federal AI safety regulation remains stalled and the EU AI Act's implementation timeline stretches to 2027, sub-national governments in the US are finding that environmental planning and resource allocation law gives them immediate, legally defensible tools to constrain AI infrastructure. The New York and Seattle moratoriums are passed under existing land-use and environmental authority — they do not require new AI-specific legal frameworks. This is significant because it means the constraint is immediate, not subject to the multi-year delays of bespoke AI regulation, and the evidentiary basis (drought data, energy grid stress) is harder to contest than risk-based AI safety arguments. Policy professionals should expect this track to accelerate: any jurisdiction with a stressed energy grid or water system now has a ready template.
The Implementation Gap Between Public Sector AI Deployment and Legal Liability Reform Is Widening Across Multiple Jurisdictions
The UK court AI pilot and the NHS liability warning are symptomatic of a broader pattern: governments are deploying AI in high-stakes public services on efficiency grounds while simultaneously failing to update the legal frameworks that govern error, harm, and accountability in those services. The EU AI Act classifies court and medical AI as high-risk and mandates conformity assessments, but does not resolve the underlying liability allocation question. The UK, operating outside the EU framework, has even less structural pressure to address this. What is emerging is a set of deployment-first, liability-later decisions that will generate the first major AI public sector negligence cases within the next two to three years — cases that will then drive reactive rather than anticipatory legal reform.
Synthetic Media Harms Are Driving Political Action Faster Than Systematic AI Governance
The xAI lawsuit by a sitting UK MP, the Prime Minister's public endorsement of that action, and the simultaneous advancement of the NO FAKES Act in the US Senate together indicate that non-consensual synthetic imagery is the policy pathway most likely to produce near-term enforceable AI legislation. This is partly because the harm is concrete and politically sympathetic, partly because it maps onto existing legal categories (defamation, harassment, intellectual property), and partly because it affects politicians directly — a dynamic that consistently accelerates legislative timelines. The risk for policy coherence is that synthetic media legislation will be passed ahead of, and potentially in tension with, broader AI liability and safety frameworks, producing a fragmented legal landscape that is harder to enforce and easier for platforms to arbitrage.
Explore Other Categories
Read detailed analysis in other strategic domains