Public Policy & Governance

Treasury Secretary Scott Bessent's decision to summon major bank chiefs — with Fed Chair Jerome Powell reportedly present — following Anthropic's release of Claude Mythos marks a qualitative shift in how US financial regulators are engaging with frontier AI. This was not a scheduled consultation or a rulemaking process; it was an emergency convening triggered by a specific commercial product launch, according to The Guardian. That framing matters: it positions advanced AI capabilities as a near-term operational risk to critical financial infrastructure, not a theoretical future concern.

This action sits in tension with the broader Trump administration posture of deregulating AI. Bessent's intervention suggests that within the administration, sectoral financial regulators — operating under existing statutory authority over systemic risk — may move faster and with more legal grounding than any new AI-specific legislation. The OFR, FSOC, and prudential regulators already have tools to mandate risk disclosures and stress-testing for novel technology risks. Separately, a DC Circuit ruling denying Anthropic's petition for a stay in an unrelated supply chain enforcement action, noted in Lawfare, suggests the courts are not providing AI companies an easy exit from regulatory obligations either.

Professor Dame Wendy Hall's testimony to the House of Commons business and trade committee — characterising China as the current 'good guy' on AI governance while describing the US approach as 'wild west' — carries institutional weight beyond parliamentary rhetoric. Hall co-authored the UK government's foundational AI review and served on the UN AI advisory board; her framing reflects the operational reality faced by multilateral governance bodies where US withdrawal from collaborative norm-setting has created a vacuum that China is actively filling, according to The Guardian.

For UK policy professionals, this testimony lands at a critical moment. The UK is navigating its post-Brexit position as a would-be AI governance hub — the Bletchley Process, the AI Safety Institute, and the proposed AI legislation all depend on transatlantic alignment that is now structurally weakened. If the parliamentary committee's findings harden into formal recommendations, the government faces a binary: continue treating UK-US alignment as the anchor of its AI governance strategy, or pivot toward EU and multilateral frameworks where China's engagement makes the political geometry more complex.

The European Commission's one-year assessment of the AI Continent Action Plan claims milestone delivery on industrial AI adoption, talent pipelines, and compute infrastructure, according to the European Commission. The Action Plan was designed as the demand-side complement to the AI Act's supply-side regulation, aiming to accelerate European industrial uptake while compliance frameworks are built out. The Commission's self-assessment predictably emphasises outputs — AI factories, testing and experimentation facilities, sectoral deployment initiatives — over the harder question of whether the AI Act's enforcement machinery is operationally ready.

The implementation gap remains significant. The AI Act's high-risk system provisions require national market surveillance authorities to be designated and resourced, conformity assessment bodies to be notified, and the AI Office to operationalise its model evaluation capabilities — none of which are complete at scale. The Action Plan's industrial momentum therefore risks running ahead of the governance architecture intended to manage its risks. This is not rhetorical: companies deploying AI in regulated sectors are making compliance investments now against regulatory frameworks whose enforcement details remain unsettled.

The convergence of Republican state legislators in Texas, progressive Democrats, and local community coalitions in opposing unregulated AI data centre construction represents a structurally significant political development, as analysed by The Guardian. The specific objections — water consumption, energy grid stress, property tax structures, and community consent — are all governed by existing permitting, utility regulation, and land use frameworks rather than any AI-specific law. This means regulatory friction is accumulating through non-AI mechanisms that are already legally operative.

For federal AI policy, this trend exposes a structural gap: the Trump administration's approach of removing federal AI oversight while accelerating infrastructure buildout has not neutralised state and local opposition; it has redirected it into existing regulatory channels that are harder to preempt. States like Texas with deregulated electricity markets face particular grid stability questions that utility commissions — not AI policy offices — will adjudicate. The practical outcome may be that AI infrastructure governance in the US is shaped more by utility regulators, county commissioners, and state environmental agencies than by any dedicated federal AI framework.

The controversy over Rosedale residents' proposal to deploy Flock's AI-powered licence plate recognition system as a private neighbourhood surveillance network — reported by The Guardian — is a concrete case study in the regulatory gap between national AI governance aspirations and local enforcement reality. Canada's proposed Artificial Intelligence and Data Act (AIDA) under Bill C-27 has stalled in parliament; Ontario has no provincial AI surveillance framework; and Toronto's municipal government has no specific authority over privately-funded, privately-operated AI systems deployed on public roads.

The Flock system is already operational in hundreds of US jurisdictions under equally inconsistent oversight — some US cities have banned it, others have licensed it, and most have neither. The governance question is not primarily about whether the technology is accurate but about who has authority to approve, audit, or prohibit private actors creating de facto surveillance infrastructure on shared public space. This is a gap that neither federal privacy law nor municipal by-law powers currently fill in Canada, and it is being forced into visibility by wealthy residents with the resources to deploy commercial AI infrastructure faster than regulators can respond.