Public Policy & Governance
Top Line
U.S. state legislatures are passing AI regulations at scale despite federal inaction, with Politico reporting that tech lobbyists are conceding defeat as rising public anger, pro-regulation funding, and Washington's paralysis embolden state-level lawmakers — a structural shift in the U.S. regulatory landscape.
Australia's parliamentary committee chairs have formally put Canberra on notice over AI-generated submissions introducing hallucinated information into federal inquiries, marking a concrete institutional integrity crisis in democratic deliberation rather than a hypothetical future risk.
The Trump administration has filed a brief siding with OpenAI against the New York Times, signalling that U.S. executive-branch AI policy now encompasses active litigation positioning on copyright — with direct implications for training data regulation globally.
New York City has formally announced a ban on AI use in public schools through eighth grade, effective immediately, making it one of the largest public education systems in the world to codify an AI restriction into enforceable district policy.
A European class action by Uber drivers across the UK and Netherlands alleges that AI-driven pay-setting systems breach GDPR, establishing a live enforcement test of whether existing data protection law can constrain algorithmic labour management.
Key Developments
U.S. State AI Regulation: The Federal Vacuum Is Being Filled from Below
The most structurally significant development in U.S. AI governance this week is the confirmation that state-level AI legislation has moved from contested to entrenched. Politico reports that technology lobbyists are effectively conceding the fight against state AI laws, citing three reinforcing dynamics: surging public distrust of AI companies, the absence of any preemptive federal framework from Washington, and a counter-mobilisation of pro-regulation funding that has neutralised Silicon Valley's traditional lobbying advantage. This is not a story about individual bills — it is a story about a durable shift in the political economy of AI regulation.
The strategic implication for multinationals is a patchwork compliance environment that increasingly resembles the pre-GDPR EU member-state landscape. Companies will face differing obligations on transparency, algorithmic accountability, and sector-specific use cases depending on which states they operate in. Senator Bernie Sanders' concurrent federal proposal to ban artificial superintelligence development, reported by Politico, is politically significant less for its legislative prospects than as a signal of how far the Overton window has shifted: a serious federal legislator is now proposing existential-level AI restrictions, and framing it as a response to documented cyberattack incidents rather than speculative risk.
Australia: AI Hallucinations Enter the Parliamentary Record — An Institutional Integrity Problem
Australia's parliamentary committee system is facing a concrete governance failure: AI-generated submissions containing factually incorrect or hallucinated content are entering the formal record of federal inquiries. Guardian Australia reports that committee chairs from multiple key committees have issued formal warnings and called on MPs to scrutinise submitted material more rigorously. This is not a hypothetical risk — it is a documented, ongoing contamination of the evidentiary basis on which legislation is developed.
The policy response gap is significant. Australia has no enacted AI-specific legislation and its voluntary AI Safety Standard, while substantive, carries no enforcement mechanism. The committee system operates on trust in submitter good faith; there is no authentication infrastructure, no detection mandate, and no penalty framework for submitting AI-generated misinformation. Comparatively, the EU AI Act includes provisions on transparency in public-sector AI interactions, but neither Australia nor any comparable Westminster system has yet legislated requirements for AI disclosure in parliamentary submissions. This incident should be understood as a stress test that existing institutional design has failed.
Trump Administration's Litigation Positioning on Copyright Sets a U.S. Policy Marker
The Trump administration's decision to file in support of OpenAI in the New York Times lawsuit is a concrete executive-branch policy action, not a rhetorical position. As reported by The Guardian, the administration's brief argues that use of copyrighted material for AI training is legally permissible — a position that, if upheld, would effectively endorse a permissive training data regime with no compensation requirement for rights holders. This directly contradicts the direction of travel in the EU, where the AI Act includes opt-out rights for copyright holders under the text and data mining provisions, and in the UK, where DSIT has faced sustained pressure from creative industries over similar questions.
The cross-jurisdictional divergence here is substantive. If U.S. courts and the executive branch cement a permissive fair-use interpretation for AI training, while the EU enforces opt-out mechanisms, the result is regulatory arbitrage: AI developers face strong incentives to train models in U.S. jurisdictions and deploy globally. This is not merely a copyright issue — it determines the economic viability of European and UK content industries and shapes where foundational AI development occurs geographically.
GDPR as an AI Labour Regulation Tool: The Uber Class Action Tests Existing Law
The multinational class action filed by Uber drivers across the UK, Netherlands, and other jurisdictions — reported by The Guardian — is a significant enforcement test of a core regulatory question: does existing GDPR provide meaningful protection against algorithmic management systems that set pay and allocate work? The claim alleges breach of data protection law, which means it does not require new AI-specific legislation to proceed. It relies on existing rights to explanation, to contest automated decisions, and to lawful basis for processing — rights that have been on the books since 2018 but rarely tested at scale against labour-facing AI systems.
The implementation gap this exposes is substantial. Regulators in both the UK (ICO) and Netherlands (AP) have issued guidance on algorithmic decision-making, but enforcement actions against live commercial systems at this scale have been rare. If the claim succeeds — and the multi-jurisdiction structure is designed to make it difficult for Uber to forum-shop — it would establish that GDPR is a de facto AI labour regulation regime, with damages exposure in the billions. This would have immediate compliance implications for any platform economy operator using AI-driven pay or task allocation systems across EU/UK jurisdictions.
Signals & Trends
Democratic Institutions Are Experiencing AI-Driven Integrity Failures Ahead of Any Governance Framework
The Australia parliamentary submissions case is not an isolated incident — it is an early indicator of a systemic vulnerability in any public consultation or legislative process that relies on written submissions. Westminster-model parliaments, EU consultations, U.S. notice-and-comment rulemaking, and UN treaty processes all operate on the assumption that submissions reflect genuine human deliberation. AI-generated content at scale breaks that assumption without breaking any current rule. Governments that are designing AI governance frameworks have not yet turned the governance lens on their own deliberative infrastructure. The first jurisdiction to mandate AI disclosure in formal public submissions and to invest in verification tooling will set a precedent that others will be under pressure to follow. The window before this becomes a systemic crisis — rather than a manageable problem — is closing.
The U.S. Regulatory Map Is Fragmenting in Ways That Will Force a Federal Response Within Two Years
The collapse of Silicon Valley's state-level lobbying strategy, combined with Sanders-style proposals at the federal level and New York City's school AI ban, points to a U.S. political environment where AI regulation is now electorally advantageous rather than toxic. This is a structural shift from 2023-2024, when federal preemption — or simple inaction — was the tech industry's preferred outcome. The emerging risk for industry is not that individual state laws are particularly onerous, but that their cumulative compliance burden becomes the forcing function for a federal framework on terms less favourable than what was on offer two years ago. Policy professionals should track which state laws include private rights of action, since those are the provisions that create durable compliance pressure independent of state AG enforcement capacity.
The UK Government's Infrastructure Visibility Gap Is a Governance Liability, Not Just an Environmental One
The FoI finding that the UK's Department for Science, Innovation and Technology does not hold information on what the country's largest datacentres are being used for — reported by The Guardian — should be read as a national security and strategic planning failure as much as an environmental oversight. A government that cannot identify who is operating critical AI infrastructure and for what purpose cannot assess concentration risk, foreign ownership exposure, or resilience dependencies. This gap is especially acute given that DSIT has since been dissolved as a ministry. The parallel planning controversy in east London over datacentre development signals that local governance frameworks are also unprepared for the infrastructure demands of AI deployment. Other G7 governments face analogous gaps, but the UK's combination of aggressive datacentre expansion policy and minimal visibility is a particularly acute version of the problem.
Explore Other Categories
Read detailed analysis in other strategic domains