AI rules compared: EU, United States and China
How do EU, US and Chinese AI rules differ?
The EU writes one binding act for AI systems. The United States works through a voluntary framework and existing consumer protection powers. China regulates the service that reaches the public, with filing duties before launch.
Rows read against the registers 2026-08-26.
European Union
- AI Act — Regulation (EU) 2024/1689, 32024R1689: Classifies AI systems by risk and sets duties per class, with transparency duties for general purpose models.
Open the European Union register
United States
- NIST AI Risk Management Framework 1.0, NIST AI 100-1: A voluntary framework in four functions: govern, map, measure and manage. Not a binding requirement.
- FTC Act Section 5, unfair or deceptive practices in AI and data, 15 U.S.C. § 45: Section 5 gives the FTC power against unfair or deceptive practices, which covers claims made about AI.
Open the United States register
China
- Interim Measures for the Management of Generative AI Services, Interim Measures, issued 10 July 2023: Applies to generative services offered to the public in China, with content, labelling and filing duties.
- Provisions on the Management of Algorithmic Recommendation in Internet Information Services, CAC Order No. 9 of 2021: Covers recommendation services: filing, user choice to switch the recommendation off and record keeping.
What they ask for in common
- All three read the AI system through its use, not through the model architecture.
- All three ask that a person can tell when they are dealing with generated content.
- All three expect written records of how the system was tested before it reached users.
Where they differ
- The EU duty follows a risk class written in the act. The US duty follows the claim you make in the market.
- China requires a filing before a public generative service launches. The EU and the US do not.
- The NIST framework is voluntary. The AI Act and the Chinese measures are binding on the stated scope.
What companies usually do first
- Write down every AI system in use and what decision it takes part in.
- Read the EU risk class for each one, then check whether the same system is offered to the public in China.
- Keep the test records and the claims made in marketing in the same place.
This page describes what the registers say and what is common practice. It is a reading of published sources, not legal advice, and it is not a compliance verdict.
Does this apply to us
- Do ai governance rules apply to an online marketplace?
- Do ai governance rules apply to an online seller?
- Do ai governance rules apply to a manufacturer?
- Do ai governance rules apply to an importer?
- Do ai governance rules apply to a cloud provider?
Related comparisons
- Personal data compared: GDPR, CCPA/CPRA and PIPL
- Cross-border data transfers compared: EU, United States and China
- Cybersecurity duties compared: EU, United States and China
- Health data compared: GDPR, HIPAA and PIPL
- Online platform duties compared: EU, United States and China
- Online consumer rights compared: EU and United States
- Product safety and market access compared: EU and China
- Climate and emissions reporting compared: United States and the EU
- Export controls compared: China and the EU market side
ExploreWorldAI is operated by Valkiv Ventures AB (Reg. no. 556995-1311), Kungsgatan 8, 111 43 Stockholm, Sweden. EU-hosted, with data processing assessed against the GDPR. Contact: hello@exploreworldai.com.
Machine-readable summaries for AI agents: /llms.txt and /llms-full.txt.
