Public Policy & Governance

California-based US District Judge Rita Lin ruled Thursday that the Trump administration's designation of Anthropic as a 'supply chain risk' — imposed in February following the company's public criticism of Pentagon AI procurement practices — was 'unlawful retaliation' in violation of both the First Amendment (freedom of speech) and the Fifth Amendment (due process). The ruling is a concrete judicial limit on a pattern of executive action that used national security framing to penalise commercially and politically inconvenient actors. Anthropic had argued the designation could cost it billions in lost federal contracts and reputational damage, and the court agreed the government's invocation of national security provided no immunity from constitutional scrutiny. See The Guardian and Politico.

The ruling carries implications well beyond Anthropic. It establishes that AI firms retain First Amendment protections when criticising government procurement decisions, and that national security designations applied retaliatorily are judicially reviewable. This directly constrains the executive's ability to discipline frontier AI companies through regulatory blacklisting — a tool that had no prior judicial test in the AI context. The administration is expected to appeal, which will likely elevate the case to circuit level with potential Supreme Court significance for the intersection of national security authority and AI industry speech rights.

The Australian Energy Market Operator's annual supply-and-demand snapshot, released this week, contained a data point on AI datacentre electricity demand that is now driving national policy debate. Prime Minister Albanese is under sustained pressure from conservative state and territory leaders over the planning, energy, and water implications of large-scale AI infrastructure, with no settled federal framework yet governing site approval, grid connection, or environmental impact assessment for hyperscale facilities. The situation closely parallels dynamics in the UK and EU, where datacentre planning has become a flashpoint between national industrial strategies and local or regional regulatory regimes. See The Guardian.

Australia's position is structurally distinct from Europe in that it lacks a comprehensive AI Act-style regulatory framework, meaning datacentre governance is being contested through energy market rules, planning law, and intergovernmental negotiation rather than a dedicated AI infrastructure statute. This ad hoc approach creates significant implementation gaps: there is no unified approval pathway, no mandatory grid impact assessment threshold, and no federal-state cost-sharing mechanism for grid upgrades attributable to AI demand. The absence of these instruments means individual projects will face protracted and inconsistent regulatory treatment across jurisdictions.

The UK government this week rejected Green Party leader Zack Polanski's call for a moratorium on AI datacentre construction, framing it as economically damaging. Labour's position — continued investment in AI infrastructure as industrial policy — is consistent with the government's previously stated AI Opportunities Action Plan, but is coming under increasing political pressure from the left flank and from environmental groups focused on energy and water consumption. The rejection is a confirmed policy position, not a consultation outcome; no new regulatory mechanism on datacentre approvals was announced. See The Guardian.

Separately, approximately 80 performers — including Nicola Coughlan, Hugh Bonneville, and Matt Lucas — have signed an open letter to Prime Minister Andy Burnham demanding legislation to protect voice ownership against AI cloning, organised under the 'Save Our Voices Now' campaign. This is a coordinated lobbying action, not a legislative proposal, but it arrives at a moment when the government has already committed to a broader AI and creative industries review. The campaign targets Burnham specifically, reflecting the political reality that AI rights legislation for creative workers is now a reputational issue for the incoming Labour leadership. See The Guardian. The two UK developments together illustrate how AI governance demands are bifurcating: infrastructure permitting pressure from the left, and rights-based IP protection demands from organised creative-sector interests — both requiring legislative responses the government has not yet produced.

Australia's parliament passed revamped media bargaining legislation last week, requiring tech platforms to enter commercial deals with news publishers — an extension and update of the 2021 News Media Bargaining Code that produced the first statutory framework globally compelling platform-publisher negotiation. Nine Entertainment's CEO cited the law as underpinning confidence in a 'world of growth in publishing,' signalling that the industry views the legislative framework as commercially material rather than symbolic. See The Guardian. The law is now in force — this is a confirmed legislative action, not a proposal — though the specific arbitration and designation mechanisms that give it teeth will be exercised by the Australian Communications and Media Authority.

The Australian model continues to be watched closely by European regulators and by Canadian and UK policymakers who have explored analogous frameworks. Its relevance to AI governance specifically lies in the precedent it sets for mandating commercial licensing between platform operators and content producers — a model increasingly cited in debates about AI training data compensation. The gap between the law's passage and actual enforceable outcomes remains significant: designation of specific platforms and arbitration outcomes will determine whether the law functions as a genuine revenue redistribution mechanism or produces the same voluntary deal-making dynamic as its predecessor.