International Regulation

International AI governance combines statutes, existing sector law, administrative programmes and voluntary guidance. These instruments differ in who they cover and what legal consequences follow. This chapter records selected developments reviewed on 7 September 2026, distinguishing enacted obligations, proposals and international commitments.

Canada — the C-27 proposal and existing obligations

Bill C-27, which included the proposed Artificial Intelligence and Data Act (AIDA), did not become law in the parliamentary session that ended on 6 January 2025. The parliamentary record shows it at House committee consideration. Its proposed obligations are not operative law, and its legislative history does not establish a date or final form for replacement legislation.¹

Assess AI deployments under the privacy, human-rights, consumer and sector laws applicable to the organisation and activity. Federal public-sector automated decisions also have their own administrative framework: the Directive on Automated Decision-Making applies within its stated government scope, rather than to every private business using AI.²

United Kingdom — existing regulators and AI Growth Labs

The government’s 21 October 2025 announcement described a proposed AI Growth Lab and a call for evidence on a cross-economy sandbox. It did not enact a comprehensive AI statute or establish that one would commence in the second half of 2026.³

The legal-services advisory AI Growth Lab subsequently launched as a programme helping participants navigate existing regulation. Government guidance updated 27 August 2026 says applications opened on 3 August and close on 27 September 2026. It expressly states that participation provides no regulatory approval, endorsement or exemption from legal obligations. Legal-services regulators and the Information Commissioner’s Office provide coordinated support.⁴

For implementation, identify applicable data-protection, online-safety, professional and sector requirements first. Treat sandbox advice and policy proposals according to their stated status. The programme’s current application timetable is not a timetable for a general AI Act.

South Korea — AI Basic Act

The AI Basic Act and its Enforcement Decree came into force on 22 January 2026. MSIT’s implementation announcement distinguishes transparency obligations, advanced-AI safety obligations, and responsibilities associated with high-impact AI.⁵

• Transparency: determine the notice and output-disclosure duties for the relevant high-impact or generative-AI service.
• Advanced-AI safety: the decree describes a separate combination of training compute, technological advancement and potential fundamental-rights impact.
• High-impact AI: assess the specified application area and risks, then the applicable operator responsibilities. Areas include recruitment, credit assessment, healthcare and education.
• Implementation period: MSIT announced a grace period of at least one year, generally deferring fact-finding investigations and penalties, with exceptional investigations for serious harms.

MSIT leads implementation of the AI Basic Act. Personal-data obligations also require assessment under the separate privacy regime. Consult current guidance for detailed applicability, including requirements affecting overseas operators; the implementation period does not erase other applicable law.

Japan — AI Promotion Act

Japan’s Act on Promotion of Research and Development, and Utilization of AI-related Technology was enacted on 28 May 2025, partly commenced on 4 June 2025, and fully commenced on 1 September 2025.⁶

The act establishes an AI Strategic Headquarters, chaired by the Prime Minister, and a national AI Basic Plan. It provides for research support, guidelines, information gathering, investigation of cases affecting rights and interests, and guidance or advice to businesses. Business operators have responsibilities to cooperate with government measures.

The statute does not establish an AI Act-style penalty schedule. That does not make the statute itself voluntary or displace existing privacy, consumer and sector law. Distinguish its legal institutional framework from the nonbinding guidance issued within it.⁶

China — AI-generated synthetic-content labels

The Measures for the Labeling of AI-Generated Synthetic Content, published 14 March 2025, became effective 1 September 2025. They apply to specified online information-service providers and build on China’s algorithm-recommendation, deep-synthesis and generative-AI instruments.⁷

The central distinction is between explicit labels, perceptible to users, and implicit labels, embedded in file data or metadata. The measures cover generated or synthesised text, images, audio, video and virtual scenes. They prescribe different duties for generation services, distribution services, app-distribution platforms and users.

Article 9 permits specified provision of content without an explicit label following agreement on user responsibilities and retention of relevant logs for at least six months. Other provisions address downstream identification and prohibit malicious removal, alteration or concealment of labels. Applicability depends on the activity and conditions in the measures. The Chinese original is controlling.⁷

Brazil — PL 2338/2023 remains a proposal

The Chamber of Deputies’ record for PL 2338/2023 shows the proposal awaiting the rapporteur’s opinion in the special committee at this review. Its 29 April 2025 entry records constitution of, and referral to, that committee; it is not a completed committee report.⁸

The Senate-origin proposal addresses AI development and use, risks and protections for individuals. Its text and timetable remain subject to the legislative process. An anticipated passage date is not an operative compliance deadline. Existing Brazilian law must be assessed independently while the proposal proceeds.

Singapore — governance and testing tools

Singapore’s AI Verify toolkit and model governance frameworks provide technical and organisational resources. The Model AI Governance Framework for Agentic AI, launched 22 January 2026, addresses bounded autonomy, meaningful human accountability, lifecycle controls and user responsibility. It complements earlier conventional-AI and GenAI guidance.⁹

IMDA subsequently announced an update incorporating additional case studies and practices. Use the current framework appropriate to the system, especially where an agent can take actions through tools. These resources support risk management; they do not establish universal safety or substitute for applicable law.¹⁰

OECD and UNESCO

The OECD AI Principles, adopted in 2019 and updated in May 2024, provide five values-based principles and policy recommendations. They address inclusive growth, human rights, transparency, robustness and accountability. They are a reference for policy and governance, not proof of compliance with every applicable law.¹¹

UNESCO’s Recommendation on the Ethics of Artificial Intelligence, adopted in November 2021, addresses human rights, dignity, fairness, human oversight and policy areas including data governance and education. Its assessment tools can help governments and organisations examine ethical implications; the recommendation should be distinguished from domestic legislation.¹²

Summit declarations and scientific reports

The Paris AI Action Summit, held 10–11 February 2025, issued the Statement on Inclusive and Sustainable AI for People and the Planet. The presidency’s published signatory list includes countries and regional organisations; the US and UK are absent from that list. Consult the current list when reporting participation, since a fixed count can become outdated.¹³

The International AI Safety Report 2025 was published on 29 January 2025, ahead of Paris. The 2026 report, published 3 February 2026, is the subsequent scientific assessment of general-purpose AI capabilities, risks and mitigation. Scientific reports and political declarations serve different purposes; neither by itself imposes domestic duties.¹⁴

India’s AI Impact Summit has already taken place: the February 2026 New Delhi meeting concluded with the New Delhi Declaration on AI Impact. The Indian government’s 21 February release records endorsements by countries and international organisations.¹⁵

Comparing instruments before applying them

Jurisdiction or instrument	Verified status or development	Practical distinction
Canada C-27/AIDA	Prior-session proposal did not become law	Assess existing law and the government directive separately
UK AI Growth Lab	Legal-services advisory programme open in August 2026	Participation does not waive law
South Korea	Act and decree effective 22 January 2026	Separate transparency, advanced-AI safety and high-impact duties
Japan	Full commencement 1 September 2025	Statutory institutions and cooperation duties; guidance is distinct
China	Labeling measures effective 1 September 2025	Actor, content type and explicit/implicit labels matter
Brazil PL 2338/2023	Awaiting special-committee opinion	Proposed text is not an operative statute
Singapore	Model frameworks and testing tools, including 2026 agentic guidance	Governance resources; legal duties remain separate
OECD, UNESCO and summit statements	International principles and commitments	Identify domestic implementation before treating them as law

For EU law, US federal policy and US state requirements, use the dedicated chapters and their dated source records.

Footnotes

1. Parliament of Canada. Bill C-27, 44th Parliament, first session. ↩

2. Treasury Board of Canada. Guide on the Scope of the Directive on Automated Decision-Making. ↩

3. UK Government. AI Growth Lab call for evidence. ↩

4. UK Government. Legal services advisory AI Growth Lab: overview, updated 27 August 2026. ↩

5. Ministry of Science and ICT. AI Basic Act implementation announcement. ↩

6. Cabinet Office of Japan. Official outline of the AI Act; Ministry of Justice. Act translation. ↩ ↩²

7. Cyberspace Administration of China and other issuing authorities. AI-generated synthetic-content labeling measures. ↩ ↩²

8. Chamber of Deputies. PL 2338/2023 official legislative record. ↩

9. IMDA. Launch of the Model AI Governance Framework for Agentic AI. ↩

10. IMDA. Updated agentic framework and deployment initiatives. ↩

11. OECD. AI Principles. ↩

12. UNESCO. Recommendation on the Ethics of Artificial Intelligence. ↩

13. Presidency of France. Paris statement and published signatory list. ↩

14. International AI Safety Report. 2025 publication; 2026 publication. ↩

15. Government of India, Press Information Bureau. New Delhi declaration and summit conclusion, 21 February 2026. ↩