Specialty Digest

Global English editionLearners’ Study Dashboard — in development
The Knowledge Library for All

AI Regulation

nounAI Ethics, Safety and Governancealso regulation of artificial intelligence, AI law, AI legislation

In one line

AI regulation is the body of binding laws and rules that governments adopt to control how artificial intelligence is developed and used, such as the European Union's risk-based AI Act, sector regulations and national measures on safety, transparency and rights.

AI regulation is the body of binding laws and rules through which governments control how artificial intelligence is developed, sold and used. It ranges from comprehensive statutes, such as the European Union’s Artificial Intelligence Act, to the application of existing consumer, privacy, discrimination and product safety laws to automated systems. Regulation is one part of wider AI governance and puts AI ethics and responsible AI into enforceable form, alongside AI accountability and AI safety.

Why regulate AI

  • Protect rights and safety. Prevent discrimination, unsafe products, manipulation and unlawful surveillance.
  • Create legal certainty. Give companies clear rules and users clear remedies.
  • Build trust. Confidence encourages adoption.
  • Level competition. Common rules reduce a race to the bottom.
  • Manage systemic risk. Address risks from very capable models and critical infrastructure uses.

The European Union’s AI Act

The Artificial Intelligence Act, Regulation (EU) 2024/1689, is the first comprehensive AI law of a major jurisdiction. It entered into force on 1 August 2024, with prohibitions and other duties applying in stages over the following years [1].

### Risk-based structure

  • Unacceptable risk. Banned practices, including social scoring by public authorities, certain manipulative techniques and some uses of biometric identification and emotion recognition in specified settings [1].
  • High risk. Permitted with strict duties; covers uses such as critical infrastructure, education, employment, essential services, law enforcement and migration. Requirements include risk management, data governance, documentation, logging, human oversight, accuracy and cybersecurity.
  • Limited risk. Transparency duties, such as telling people they are talking to a chatbot and labeling synthetic content; see AI transparency.
  • Minimal risk. Most systems, with no new obligations.

### General-purpose models

Providers of general-purpose AI models face duties on documentation, copyright policy and training data summaries, with extra obligations for models judged to pose systemic risk, such as evaluations, incident reporting and cybersecurity.

### Enforcement

National authorities and a European AI Office supervise the rules, and penalties can reach a percentage of worldwide turnover for the most serious violations. Dates, guidance and standards continue to develop, so readers should consult current official texts.

Other approaches around the world

  • United States. No single federal AI statute as of this writing. Policy relies on sector regulators, existing laws, executive actions, voluntary frameworks such as the NIST AI Risk Management Framework and a growing number of state laws on topics like hiring tools and deepfakes.
  • China. Targeted rules on recommendation algorithms, synthetic media and generative AI services, with requirements on registration, content and security review.
  • United Kingdom. A principles-based, regulator-led approach with an AI Security Institute to evaluate advanced models.
  • Canada, Brazil, Japan, South Korea and others. Proposed or adopted laws and guidelines at differing stages, some based on risk.
  • International. The Council of Europe’s 2024 Framework Convention on AI is the first legally binding international treaty in this area [2]; the OECD, UNESCO, the G7 and the United Nations issue principles and coordination statements.

Laws change quickly, so any summary is a snapshot; see the OECD’s policy observatory for tracking [3].

Related laws that already apply

  • Data protection. Rules such as the EU’s GDPR; see AI privacy.
  • Anti-discrimination and consumer protection. Applied to automated decisions; see algorithmic bias.
  • Product safety and liability. For AI embedded in devices and vehicles.
  • Intellectual property. Ongoing cases about training data and generated works.
  • Sector rules. Medical devices, financial services, aviation and others.

Debates

  • Innovation versus protection. Critics warn that heavy compliance burdens could slow startups; supporters say clear rules help markets.
  • Defining AI and risk. Boundaries between categories are contested.
  • Regulating models or uses. Some favor rules on applications, others on the most capable models.
  • Open-source treatment. How far exemptions should go.
  • Fragmentation. Different national rules raise costs and complicate compliance.
  • Enforcement capacity. Regulators need technical expertise.

Preparing for compliance

Organizations typically inventory their systems, classify them by risk, assign owners, document data and testing, build in human oversight and monitor after release; see AI governance and model deployment. Legal advice is needed for specific cases.

A worked example

A company based outside Europe sells software that screens job applicants and offers it to employers in several EU countries. Under the AI Act, hiring tools fall in the high-risk category, so the provider must set up a risk management process, check that training data are relevant and as free of errors as possible, keep technical documentation and logs, design the tool so that human reviewers can understand and override its scores and test for accuracy and robustness. It must complete a conformity assessment before placing the product on the market and register the system in an EU database. Employers who use the tool must follow instructions, assign trained staff to oversee it, monitor results and inform candidates that AI is used. If the company ignores these duties, national authorities can order changes or withdrawal and impose fines. The example shows that regulation reaches both providers and users and applies across borders when systems are used in the market.

Timeline and phases of application

The AI Act does not apply all at once. Bans on prohibited practices and rules on AI literacy applied first, followed by duties for general-purpose models, and later by most requirements for high-risk systems, with additional time for systems built into products already covered by other safety laws. The European Commission issues guidelines, codes of practice and harmonized standards to explain how to comply, and national authorities set up testing environments called regulatory sandboxes where firms can try products under supervision. Because schedules, guidance and proposals to simplify or delay parts of the law can change, organizations should follow official announcements instead of relying on a fixed calendar.

Effects on business and research

Compliance has costs, especially for small firms, but it also gives clearer rules for buyers and investors. Large providers may adopt EU-style documentation worldwide, a pattern sometimes called the Brussels effect, as happened with data protection law. Research use is largely exempt, and open-source components receive some relief, though not for prohibited or high-risk uses. Standards bodies, auditors and legal advisers are growing in importance, and many companies now hire staff to track rules in each market and connect them to technical practices such as testing, logging and AI observability.

FAQs

What is AI regulation? It is the set of binding laws that governments use to control how AI is developed and used.

What is the EU AI Act? A European regulation that treats AI by risk level, bans some practices and sets strict duties for high-risk systems.

Is AI regulated in the United States? Mostly through existing laws, sector regulators, voluntary frameworks and state laws rather than one federal AI statute.

Does AI regulation apply to companies outside the EU? The AI Act can apply to providers and deployers outside the EU when their systems are placed on the EU market or their outputs are used there.

What is the difference between AI regulation and AI governance? Regulation is binding law. Governance is the broader system including standards and company policies.

More in AI Ethics, Safety and Governance

Key facts

  • The European Union's Artificial Intelligence Act, Regulation (EU) 2024/1689, entered into force on 1 August 2024, with obligations phased in over several years1.
  • The Act sorts AI systems into unacceptable, high, limited and minimal risk categories and bans certain practices such as social scoring by public authorities1.
  • The Council of Europe adopted the Framework Convention on Artificial Intelligence in 2024, the first international legally binding treaty on AI2.

Related concepts

Quick checkWhat is AI regulation?Show answer

AI regulation is the body of binding laws and rules that governments adopt to control how artificial intelligence is developed and used, such as the European Union's risk-based AI Act, sector regulations and national measures on safety, transparency and rights.

Sources

  1. European Union, Regulation (EU) 2024/1689 (Artificial Intelligence Act), https://eur-lex.europa.eu/eli/reg/2024/1689/oj
  2. Council of Europe, Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law, https://www.coe.int/en/web/artificial-intelligence/the-framework-convention-on-artificial-intelligence
  3. OECD, AI Policy Observatory, https://oecd.ai/en/

Editorially reviewed by Specialty Digest Editorial TeamLast reviewed November 17, 2026Researched and drafted with AI assistanceReport an issue