AI Regulation World Map

Brazil's AI framework law establishes a risk-based regulatory model inspired by the EU AI Act but adapted to Brazilian context. It mandates algorithmic impact assessments for high-risk systems and creates a national supervisory authority.

China has enacted multiple binding AI regulations targeting specific modalities: algorithmic recommendation, deepfakes/synthetic media, and generative AI. The Interim Measures for Generative AI require security assessments, content labelling, and data provenance for public-facing GenAI services.

The EU AI Act is the world's first comprehensive, binding AI law. It classifies AI systems by risk tier (unacceptable, high, limited, minimal) with corresponding obligations. Prohibitions on social scoring and real-time biometric surveillance in public took effect Feb 2025; full compliance required by Aug 2026.

South Korea's AI Basic Act establishes a framework for classifying high-risk AI, mandating transparency disclosures and impact assessments. It also creates an AI Committee for governance oversight.

Australia released a Voluntary AI Safety Standard with 10 guardrails in Sep 2024 and is consulting on making some guardrails mandatory. No binding AI-specific law yet, but ACCC and OAIC enforce existing laws against AI harms.

AIDA was part of the Digital Charter Implementation Act (C-27). It would have created obligations for high-impact AI systems. The bill died when Parliament was prorogued in Jan 2025. A revised AI bill is expected.

Japan favours a non-binding, agile governance approach. The AI Guidelines for Business provide voluntary principles aligned with the Hiroshima AI Process. Sector-specific binding rules (e.g., financial AI) are progressing.

The UK has adopted a sector-specific, pro-innovation approach rather than a single AI law. Existing regulators (FCA, Ofcom, CMA, ICO) apply AI principles within their domains. The AI Safety Institute tests frontier models. The new Labour government has signalled binding legislation may follow.

Kenya's Blockchain & AI Task Force recommended an AI governance framework in 2023. The Data Protection Act applies to automated decision-making. An AI-specific bill is expected.

Multiple AI regulatory bills have been introduced in the Mexican Congress. No comprehensive AI law has been adopted. Data protection law (LFPDPPP) applies to automated processing.

No federal AI law exists. EO 14110 established safety reporting requirements but was partially revoked in Jan 2025. NIST's AI Risk Management Framework is voluntary. Over 35 states have introduced AI bills, with Colorado SB 24-205 (high-risk AI disclosure) the most advanced.

Argentina has a National AI Plan and guidelines for AI in public services but no binding AI-specific legislation. Existing data protection (PDPA) and consumer protection laws apply.

India has published voluntary responsible AI principles and is developing the Digital India Act which may include AI provisions. Current approach is industry-led self-regulation with sector-specific interventions (RBI for financial AI).

Israel's Innovation Authority leads AI policy. Voluntary ethical principles for public-sector AI exist. The approach favours innovation-first regulation.

NITDA (National Information Technology Development Agency) has published a draft AI governance framework. Nigeria's National AI Strategy is in development. No binding AI law yet.

Rwanda published a National AI Policy in 2023, positioning itself as a regional AI governance leader in East Africa. Focus on responsible AI for development.

SDAIA (Saudi Data & AI Authority) has published AI ethics principles and a governance framework. Vision 2030 positions AI as a national priority. No binding AI-specific legislation yet.

Singapore's Model AI Governance Framework is voluntary but widely adopted in ASEAN. AI Verify is an open-source testing toolkit for AI governance. No binding AI-specific legislation; regulation is sector-based.

South Africa's DCDT published an AI policy discussion paper in 2024 seeking public comment. POPIA (data protection) applies to AI processing personal data. No AI-specific binding law.

Switzerland has voluntary AI guidelines for the federal administration and data protection authority guidance. No AI-specific binding law, but closely monitoring EU AI Act for equivalence.

The UAE was the first country to appoint a Minister of State for AI (2017). The National AI Strategy 2031 sets development goals. Voluntary ethics guidelines exist; no binding AI-specific law.