Daily Junction.mobi portalNews · October 2, 2025 · Amelia Hart · 5 min
The EU AI Act entered into force in August 2024 with prohibited-practice rules applying from February 2025. The US rescinded its main federal AI executive order in January 2025. The UK still has no dedicated AI law. Here is the diverging global picture.
Eighteen months ago, global AI governance looked like it might converge around a shared safety-first framework, built on the momentum of the UK's Bletchley Park summit in November 2023 and South Korea's Seoul summit in May 2024. By early 2025, that convergence had visibly cracked. The EU pressed ahead with binding, phased legislation. The US pivoted sharply toward deregulation under a new administration. The UK continued its distinctive non-statutory path. Understanding why requires looking at each jurisdiction's actual 2024-25 moves rather than their earlier stated intentions.
The European Union's AI Act — the world's first comprehensive, binding AI law — entered into force on 1 August 2024, following political agreement in December 2023. Rather than applying all at once, its obligations phase in over roughly two years: bans on "unacceptable risk" practices, including certain biometric surveillance and social scoring uses, applied from February 2025; obligations on providers of general-purpose AI models — covering transparency, technical documentation and copyright compliance — applied from August 2025; and most remaining high-risk system requirements phase in through August 2026. Non-compliance penalties can reach up to 7% of a company's global annual turnover for the most serious violations.
The United States took the opposite trajectory. President Biden's October 2023 executive order, EO 14110, had directed federal agencies to develop AI safety testing, reporting and risk-management requirements — the closest the US had come to a coordinated federal AI governance framework. On taking office in January 2025, President Trump rescinded that order and replaced it with a policy emphasising reduced regulatory burden and US competitiveness against China's AI development, leaving oversight to existing sector regulators — the FTC on consumer harms, the FDA on medical AI, financial regulators on AI in banking — rather than a dedicated cross-government approach.
| Jurisdiction | Framework | Key 2024-25 milestone | Approach |
|---|---|---|---|
| European Union | AI Act (binding law) | Entered into force Aug 2024; phased through Aug 2026 | Risk-tiered, comprehensive, binding |
| United States | No federal AI statute | Biden's EO 14110 rescinded Jan 2025 | Sector-by-sector, deregulation-leaning |
| United Kingdom | 2023 AI white paper (non-statutory) | AI Safety Institute renamed AI Security Institute, 2025 | Existing regulators apply existing powers |
The UK has held its position closest to the status quo. Its 2023 AI white paper deliberately avoided creating a single AI statute or regulator, instead asking existing bodies — the Information Commissioner's Office, Ofcom, the Financial Conduct Authority and others — to apply their existing powers within their sectors. The UK's AI Safety Institute, established after Bletchley Park, was renamed the AI Security Institute in 2025, a change widely read as reflecting a shift in framing from broad AI "safety" toward more narrowly defined national security risks.
The clearest sign of fracture came at the Paris AI Action Summit in February 2025, the third in the Bletchley-Seoul-Paris summit series. Most attending nations signed the summit's joint declaration on AI development principles; the US and UK were notable non-signatories, both citing concerns that the declaration's language on governance and safety was insufficiently focused on competitiveness or precision. That represented a visible retreat from the more unified tone of the 2023 Bletchley Park declaration, which the US, UK, EU and China had all signed.
"Two years ago the conversation was almost entirely about existential risk and shared safety standards. By 2025 it had become as much about industrial policy and competitive positioning as about safety." — a shift widely noted across coverage of the Paris summit compared with Bletchley Park and Seoul.
For any UK business selling AI-enabled products or services into the EU — a large share of UK tech exporters — the EU AI Act's extraterritorial reach means compliance timelines matter regardless of where the company is based: if your AI system is used by EU customers, the Act's phased obligations generally apply, with the general-purpose model rules already live since August 2025 and high-risk system rules following through 2026. UK-only businesses face a lighter but less certain regime, since existing UK regulators are still developing sector-specific AI guidance rather than working from a single rulebook, meaning the practical compliance bar can vary significantly depending on which regulator — the ICO for data protection-adjacent AI use, the FCA for financial services AI — has jurisdiction over your specific use case. For a UK-specific angle on this, see our coverage of what artificial general intelligence actually means and how close current systems are to it.
Beyond the three jurisdictions covered in detail here, a wider patchwork is emerging that further complicates the picture for any genuinely global AI company. India, Brazil, Japan, South Korea and several Gulf states have each begun developing distinct AI governance approaches through 2024-25, generally borrowing selectively from the EU's risk-based framework, the US's lighter-touch sector approach, or entirely bespoke national priorities — meaning the fragmentation described here between three major jurisdictions is, in practice, a simplified version of an even more fractured global picture, a dynamic our coverage of the UK's own approach to artificial general intelligence policy touches on from a different angle.
Watch the EU AI Act's August 2026 full-implementation deadline as the next major compliance cliff-edge, and watch whether the US under the current administration introduces any federal AI legislation at all, or continues relying entirely on existing sector regulators and state-level rules — several US states, including Colorado and California, have moved on their own AI legislation in the regulatory gap left by the rescinded federal executive order. Whether the UK eventually moves from its non-statutory approach toward a dedicated AI law, something successive UK governments have so far resisted, will likely depend on how enforceable the EU framework proves in practice, and whether UK businesses face material competitive disadvantage from regulatory uncertainty relative to EU-based rivals operating under Europe's now-binding rules.
![[View image]](/img/ai-regulation-around-world.jpg){kind=link}