Public Policy & Governance

President Trump announced the creation of an 'AI Force' without prior consultation with Congress or, reportedly, with the tech industry — both of which were caught off-guard by the announcement, according to Politico. The announcement has no accompanying legislation, no disclosed budget line, and no clear statutory basis, meaning it currently exists as executive intent rather than enforceable policy. This follows a pattern established with Trump's earlier AI executive orders: announce boldly, operationalise slowly, and leave Congress to react rather than lead.

The practical consequence is a governance gap. Congress controls appropriations and, ultimately, the legal framework within which any new AI-related military or civilian body would operate. Without bipartisan legislative buy-in — currently absent — the 'AI Force' risks becoming an unfunded mandate or a rebranding of existing programmes. Senior members of relevant committees have not been briefed, which further constrains the administration's ability to build durable institutional infrastructure. The episode illustrates that the US continues to make major AI governance decisions through executive action alone, creating fragility and reversibility that other jurisdictions — notably the EU with its legislated AI Act — have deliberately designed out of their frameworks.

OpenAI has submitted recommendations calling for the US to lead an international push for common AI evaluation methodologies and governance mechanisms for self-improving systems, as reported by Politico. The proposal aligns with the company's broader regulatory strategy: advocate for international standards that, in practice, mirror the evaluation frameworks OpenAI already uses internally, thereby shaping compliance costs in ways that favour incumbents with established safety teams over new entrants.

The timing is significant. The UN's top tech envoy, Amandeep Singh Gill, recently gave a candid assessment of what the UN can realistically achieve on AI safety — characterised by Foreign Policy as a 'scorpions in a bottle' dynamic among competing state interests. Into this multilateral uncertainty, OpenAI is pushing for US-led bilateral and plurilateral standard-setting, which would structurally sideline the EU AI Act's conformity assessment regime and China's own algorithmic governance framework. Policy professionals should treat OpenAI's 'multilateralism' framing critically: the company is not calling for a binding international treaty but for US primacy in setting voluntary benchmarks, which is a materially different governance outcome.

Chatham House Senior Research Fellow Isabella Wilkinson gave evidence to the House of Lords International Relations and Defence Committee on September 21, covering global AI governance and systemic risks, according to Chatham House. The committee's focus on international dimensions — rather than purely domestic regulation — indicates that the Lords are building a case for the UK to take a more assertive multilateral stance, likely as a precursor to recommendations that could pressure the government to move beyond its current pro-innovation, light-touch posture.

Separately, Liberal Democrat leader Ed Davey used his party conference keynote to call for a global nuclear-style AI non-proliferation treaty to halt superintelligent AI development, as reported by The Guardian. This is political rhetoric from an opposition party with no current path to government, not actionable policy. However, it is a signal of how the Overton window is shifting in UK politics: positioning the Conservatives and Labour as insufficiently precautionary on existential AI risk is now a viable electoral wedge strategy, which will eventually constrain the governing party's room to remain purely pro-growth on AI regulation.

The Albanese government is actively considering granting AI companies broad rights to train on Australian-hosted content, effectively creating a permissive regime that would make Australia an attractive jurisdiction for AI development relative to the EU's stricter copyright framework, according to The Guardian. The reforms are framed as enabling Australia to attract AI investment and maintain relevance in the global AI economy. The creative sector, represented by organisations including musician and activist Holly Rankin's advocacy work, is mounting organised opposition — arguing that the reforms amount to government-sanctioned expropriation of creators' intellectual property without compensation.

The policy trade-off here is structurally similar to debates in the UK and EU: governments seeking AI competitiveness advantage are being pressured to weaken copyright protections that creators and publishers rely on. Australia's proposed approach would go further than the EU's text-and-data mining exception (which requires rights-holders to opt out) by granting affirmative access to open internet content. If enacted, it would likely trigger WTO-level questions about national treatment obligations and could set a precedent that other jurisdictions — particularly in the Indo-Pacific — adopt to compete for AI investment.

Citizens Advice has formally called on regulators overseeing essential services — energy, banking, telecoms, and internet — to require providers to guarantee customers access to a human agent, as reported by The Guardian. The organisation's research found that AI chatbot systems caused harm — defined as wasted time, stress, and delayed problem resolution — for more than half of users accessing essential services. This is a concrete, sector-specific policy demand directed at named regulators with existing statutory consumer protection mandates: Ofgem, the FCA, and Ofcom.

The significance for governance professionals is that this demand is actionable within current regulatory frameworks. Ofgem, the FCA, and Ofcom already have powers to set customer service standards; a Citizens Advice report of this nature is routinely incorporated into regulatory consultations. The FCA's Consumer Duty, which came into force in 2023 and requires firms to deliver good outcomes for retail customers, is a plausible existing statutory hook for enforcement action against firms whose AI customer service demonstrably causes consumer harm — without requiring new legislation.