International Regulation

Outside the EU and US, the international AI regulatory landscape moved substantially during 2025. South Korea enacted a binding AI Basic Act; Japan passed a soft-law AI Promotion Act; China brought synthetic-content labelling rules into force; Brazil continues to advance PL 2338; the UK published a Blueprint for AI Regulation with a comprehensive statute signalled for late 2026; and Canada’s Bill C-27 (which would have created the Artificial Intelligence and Data Act, AIDA) died on the order paper.

Figure: International AI regulation status as of May 2026. Enforcement maturity, not legitimacy — voluntary regimes (right column) heavily influence binding regimes (left).

Canada — AIDA effectively dead

Canada’s Bill C-27, which contained the proposed Artificial Intelligence and Data Act (AIDA), died on the order paper on 6 January 2025 when Parliament was prorogued.^[1] Minister of Innovation, Science and Industry Solomon confirmed in June 2025 that the bill will not be reintroduced in its previous form.

Canada therefore has no comprehensive horizontal AI statute. AI-related obligations continue to flow from existing federal law (PIPEDA, the Privacy Act, the Canadian Human Rights Act) and from provincial law (notably Quebec’s Law 25). The federal Directive on Automated Decision-Making continues to apply to federal government use of AI. A new approach to AI legislation is anticipated but not currently scheduled.

United Kingdom — Blueprint published, statute pending

The UK has not yet enacted a comprehensive AI statute. On 21 October 2025, the government published the Blueprint for AI Regulation, an updated policy document setting out a context-specific, sector-led approach implemented through existing regulators (Ofcom, ICO, FCA, MHRA, CMA).^[2] The Blueprint also launched an AI Growth Lab regulatory sandbox.

A comprehensive AI Bill is signalled for H2 2026. Until then, AI obligations are layered: data-protection obligations under UK GDPR and the Data Protection Act 2018, online safety obligations under the Online Safety Act 2023, sector-specific obligations from individual regulators, and copyright/IP rules applied case-by-case.

The UK AI Security Institute (formerly AI Safety Institute) remains the focal point for frontier-model evaluation and is the lead UK participant in the International Network of AI Safety Institutes.

South Korea — AI Basic Act (effective January 2026)

Korea enacted the Act on Promotion of the AI Industry and Framework for Establishing Trustworthy AI (“AI Basic Act”) in January 2025, with an effective date of 22 January 2026.^[3] Major obligations:

• Three regulatory tracks: a transparency track for generative AI interactions, a frontier-safety track for “high-impact” AI, and a governance track applied broadly.
• Extraterritorial application to providers serving Korean users.
• A one-year administrative-fine grace period following the effective date, during which violations are noted but not fined.
• Disclosure obligations for AI-generated content, similar in structure to EU AI Act Article 50 but with Korean-specific notice requirements.

The Personal Information Protection Commission (PIPC) and the Ministry of Science and ICT share enforcement responsibility.

Japan — AI Promotion Act (effective June 2025)

Japan’s AI Promotion Act passed the Diet on 28 May 2025, with provisions effective 4 June 2025 and full operation from September 2025.^[4] The Act takes a soft-law approach:

• No penalties for non-compliance.
• Establishes an AI Strategy Council at the cabinet level.
• Encourages voluntary risk management and transparency.
• Aligned with the OECD AI Principles and Hiroshima AI Process outputs.

Japan’s approach reflects a deliberate choice for an industry-friendly framework that emphasises promotion over prohibition, leaving harder regulation to sectoral regulators and existing law (Personal Information Protection Act, Telecommunications Business Act).

China — synthetic content labelling and ongoing regulation

China’s Measures for the Labeling of AI-Generated Synthetic Content, issued 14 March 2025, took effect on 1 September 2025.^[5] Providers of AI services must:

• Apply explicit labels (visible to users) to AI-generated text, images, audio, and video.
• Apply implicit labels (embedded in metadata) to enable downstream identification.
• Distinguish between labels for AI-generated content and labels for AI-synthesised content (e.g., deepfakes).

These measures build on China’s earlier instruments — the 2022 Internet Information Service Algorithmic Recommendation Management Provisions, the 2023 Provisions on the Administration of Deep Synthesis, and the 2023 Interim Measures for the Management of Generative AI Services — that collectively form one of the most comprehensive AI regulatory regimes globally.

Brazil — PL 2338/2023

Brazil’s PL 2338/2023 (“the Brazilian AI Bill”) was approved by the Senate in December 2024 and is under review by the Chamber of Deputies, with a special committee report dated 29 April 2025.^[6] The bill takes a risk-based approach broadly modelled on the EU AI Act, with high-risk categories, fundamental-rights protections, and an enforcement body. Final passage is anticipated during 2026 but not confirmed.

Singapore

Singapore’s AI Verify framework (Infocomm Media Development Authority and the AI Verify Foundation) continues as a voluntary technical testing toolkit and governance framework. Singapore is a leading participant in the International Network of AI Safety Institutes and has signed bilateral AI cooperation agreements with multiple jurisdictions.

The Model AI Governance Framework for Generative AI (2024) remains the principal Singaporean guidance for industry.

OECD AI Principles

The OECD AI Principles, originally adopted in 2019, were updated on 9 May 2024 to address generative AI and frontier-model issues.^[7] The principles — inclusive growth, human-centred values, transparency, robustness, accountability — remain the lingua franca for high-level AI policy and are incorporated by reference into numerous national instruments (including Japan’s AI Promotion Act).

UNESCO Recommendation on AI Ethics

The UNESCO Recommendation on the Ethics of Artificial Intelligence (November 2021) remains the principal multilateral ethics instrument. UNESCO continues capacity-building work with member states; the Recommendation is most influential as a reference for national ethics frameworks in jurisdictions without binding statutes.

Paris AI Action Summit and the International Network

The Paris AI Action Summit (10-11 February 2025) produced:

• A Statement on Inclusive and Sustainable AI signed by 58 countries (the United States and the United Kingdom did not sign).
• An International AI Safety Report authored by 96 international experts.
• The launch of Current AI, an international public-interest foundation with approximately $400 million in initial commitments.
• The ROOST (Robust Open Online Safety Tools) initiative.

The International Network of AI Safety Institutes continues to coordinate frontier-model evaluation between participating institutes — UK AISI, US CAISI, Singapore, Japan AISI, Korea, France INESIA, and others — despite the AISI-to-CAISI rebrand. India is hosting the next major AI summit in the series.

Regional posture summary