Global Approaches to AI Regulation: A Comparative Overview

A Global Benchmark

In 2019, the G20 adopted the OECD AI principles to guide governments and organisations in developing sustainable, human-centric approaches to artificial intelligence. These principles emphasise safeguarding human rights and the rule of law, ensuring AI's benefits are widely distributed, and requiring organisations to maintain accountability mechanisms.

Emerging policy across G20 jurisdictions reflects these principles. Canada’s pending legislation proposes mandatory requirements for high-risk AI applications, including hiring decisions and national security. These align with the OECD focus on accountability and robustness. President Biden’s Executive Order of 30 October 2023 mirrors similar priorities, including safety, security, investment in education and a commitment to civil rights and workers’ rights.

Although specific policies vary, ranging from voluntary guidance in the UK, Japan and Singapore to mandatory rules in the EU’s AI Act, the OECD principles remain central to global thinking.

A Risk-Based Approach

Governments increasingly favour a risk-based model over prescriptive rules. This approach identifies the risks posed by specific AI activities and tailors regulation to protect core OECD principles such as privacy, transparency, security and non-discrimination. The aim is to introduce oversight without stifling innovation and to keep compliance costs proportionate to the harms being prevented.

The EU’s AI Act and Canada’s AI and Data Act exemplify this model. Both use risk and impact assessments to categorise AI systems and assign compliance obligations based on risk level. Higher risk categories attract more stringent requirements. This structure aims to keep rules up to date in a rapidly evolving environment.

The EU has included one notable exception. Despite adopting a risk-based framework, the AI Act contains mandatory prohibitions on certain uses of AI across the Union, deeming them too dangerous even at this early stage.

Growing Momentum for Consensus

Beyond the G20’s adoption of shared principles, international cooperation has accelerated. In October 2023, the G7 agreed on International Guiding Principles on Artificial Intelligence and a Code of Conduct for AI developers to promote safety and trustworthiness. The United Nations has also advanced collaboration through UNESCO’s Recommendations on the Ethics of AI, adopted by countries including China, Egypt, India, Peru, Saudi Arabia, Singapore and South Korea.

In November 2023, the UK hosted the AI Safety Summit, attended by the US, China and several African nations. The summit produced the Bletchley Declaration, which commits to unlocking AI’s potential while restricting unsafe uses.

Jurisdictional Developments

European Union

The EU has taken the most significant step in AI regulation. In December 2023, the AI Act completed the political trilogue stage, with agreement reached between the European Commission, Council and Parliament. The Act is the most comprehensive AI regulation to date and appears intended to set a global benchmark similar to the EU’s success with GDPR.

The Act creates harmonised rules for placing AI on the EU market and applies to Member States and any third-country providers operating within the EU. It is sector-agnostic, regulating AI use in principle rather than by industry.

The Act categorises AI systems by risk and imposes requirements accordingly. The most stringent obligations apply to general-purpose AI systems, including generative AI, which may pose systemic risk. These obligations include human oversight, reliable documentation, risk and quality management systems and robust cybersecurity measures. Minimal-risk categories require initial assessments and ongoing transparency, and the Act encourages voluntary codes of conduct.

The Act also prohibits certain AI systems deemed manifestly injurious to vital rights, including facial recognition databases similar to those used by Clearview AI in the US.

A proposed AI Liability Directive remains under consideration. It would provide financial compensation for individuals harmed by AI technology and may set another global benchmark.

China

China was among the first jurisdictions to implement AI regulations and is a party to the OECD principles. It participated in the UK’s AI Summit and adopted UNESCO’s Recommendations on the Ethics of AI.

Current Chinese regulation is piecemeal and has emerged in response to growing concerns. Early rules targeted recommendation algorithms used in social media and navigation. More recent policies address autonomous vehicles, medical AI and facial recognition.

Because Chinese policymaking is opaque, it is difficult to determine the full scope of existing rules. Lawmakers are reportedly drafting a comprehensive AI regulation to consolidate current measures and introduce new policies. It remains unclear whether China will align with Western developments or adopt a more individualistic approach.

Canada

Canada is pursuing a stricter model similar to the EU. Its anticipated AI and Data Act aims to ensure high-impact AI systems meet safety and human rights expectations and to prohibit reckless or malicious uses of AI. The government has also published a code of practice for generative AI developers.

A Directive on Automated Decision Making issued in 2023 imposes requirements on federal government use of automated systems. Ontario is considering whether employers should disclose AI use in hiring processes, which could be a significant development in employment law.

Japan

Japan’s National AI Strategy adopts agile governance. The government provides non-binding guidance and relies on voluntary self-regulation by the private sector. This approach is reflected in publications such as the AI Utilisation Guidelines and the draft AI Operator Guidelines issued in 2023.

Like the UK, Japan aims to foster innovation and position itself as a global AI hub.

United States

US regulation began to take shape after President Biden’s October 2023 Executive Order, which called for greater transparency and new standards. The US appears likely to adopt a sector-specific model, with agencies creating their own rules.

A risk-based grading system for AI use is expected, similar to the EU’s framework. The National Institute of Standards and Technology has already proposed a model for agencies to implement.

Future progress may depend on the outcome of the upcoming election, with a second Trump administration potentially using AI policy amid broader trade tensions with China.

United Kingdom

In February 2024, the UK Government published its response to the 2023 White Paper on AI regulation. The pro-innovation messaging remains unchanged. The UK has adopted an outcome-based framework underpinned by five principles: safety, security and robustness; appropriate transparency and explainability; fairness, accountability and governance; and contestability and redress.

Existing regulators will apply current laws and issue supplementary guidance by 30 April 2024. The Government will also publish voluntary safety and transparency measures. While legislative amendments are envisaged, they are not currently considered necessary.

The UK's challenge is aligning its approach with global industry trends. Ambiguity may deter businesses from establishing themselves in the UK. Divergence from the EU may also be counterintuitive for companies operating across the 28 Member States.

Conclusion

Across major jurisdictions, 2024 is likely to be the year AI regulation becomes firmly established. Consensus is gaining momentum, with further AI Safety Summits planned in South Korea and France. However, each jurisdiction will continue to develop its own rules as governments seek to balance innovation with effective safeguards. Global collaboration will be essential, as coherent and comprehensive governance is needed to address the complex ethical and regulatory challenges posed by artificial intelligence.