AI regulation
The totality of laws and mandatory regulations that define how artificial intelligence may be developed, sold, and used.
In plain terms
AI regulation is to artificial intelligence what traffic laws are to driving. It does not ban you from driving, but tells you how fast you can go, when a license is required, and who is liable if something goes wrong.
Its goal is twofold: to protect people from harm (such as unfair decisions, deception, surveillance) while providing predictable rules for companies. Those who know what is allowed develop more boldly.
The world is not uniform on this. The European Union has enacted comprehensive legislation, the AI Act. The United States has no comprehensive federal AI law: sectoral rules and the laws of individual states (such as Colorado or California) apply, while the federal government tends to favor lighter regulation. China regulates specific areas with separate decrees, such as generative AI and the labeling of AI-generated content.
Regulation goes beyond voluntary ethical guidelines: authorities can enforce compliance, and violations incur fines.
How it works
- First, they determine what counts as an AI system and who the rules apply to: the developer, the distributor, and also the company or authority that uses it.
- Most modern regulations classify by risk: the greater the potential harm an application can cause, the stricter the rules that apply to it. A music recommender is almost unrestricted, while a credit scoring or admissions system is not.
- Some uses are banned outright. In the EU, examples include the social scoring of people and systems that manipulate or exploit vulnerabilities.
- High-risk systems come with obligations: documentation, testing, human oversight, and logging. There is also a transparency rule: you must know if you are speaking to a machine or if content was created by AI.
- Authorities monitor compliance and can impose fines. In the EU, in the most serious cases, the fine can reach 35 million euros or 7 percent of the company's annual global turnover, whichever is higher.
An everyday example
You apply for a job, and an AI system screens your CV first. Under EU rules, this is a high-risk use. As soon as the relevant regulations become applicable, the company must ensure human oversight and inform you that AI is involved in the decision.
Who is affected
A common misconception
Limits and risks
- Technology changes faster than the law: by the time a rule takes effect, new tools already exist, such as autonomous agents, for which the text does not always provide a clear answer.
- A rule is only as good as its enforcement: if the authority lacks staff and funding, strict regulations on paper remain weak in practice.
- Rules differ by country, so what is banned in the EU may be allowed elsewhere, and a foreign service may not necessarily meet local expectations.
- Deadlines and detailed rules are still evolving, so for specific cases, always check the current official text, not an older article.
Similar, but not the same
- AI ethics – Ethics is about what would be right, and ethical guidelines are mostly voluntary. Regulation, however, is mandatory and enforceable with fines.
- EU AI Act (MI-rendelet) – The AI Act is a single specific EU law, while AI regulation is a collective term for all such laws and requirements worldwide.
- AI governance – Governance is broader: it includes companies' internal regulations, standards, and international agreements, not just state laws.
Related concepts
AI governance, technology regulation, Digital law
EU AI Act, Risk-based approach, high-risk AI system, tiltott AI-gyakorlatok, transparency obligation
AI ethics, data protection (GDPR), AI safety, algorithmic bias, copyright and AI
A brief history
The first widely accepted international principles were issued by the OECD in 2019, but these were not yet mandatory. The European Commission proposed the AI Act in 2021. The regulation was adopted in 2024 and entered into force on August 1, 2024, as the world's first comprehensive AI law. Also in 2024, the Council of Europe adopted the Framework Convention on Artificial Intelligence, which is the first binding international AI treaty.
Where things stand
The EU AI Act enters into force in phases. The bans have been in force since February 2025, and the rules for general-purpose models since August 2025. The requirements for high-risk systems were originally to be applied from August 2026. However, at the end of 2025, the European Commission proposed a postponement in the so-called digital omnibus package, roughly until the end of 2027, or August 2028, so the valid deadline must always be checked in the current text. Globally, more and more countries are enacting their own rules, but there is no unified international system.
In the news
- Hungarian tech-regulation expert: EU AI Act is over-regulatory but here to stayNov 6, 2026 · Jog
- AI liability in insurance: AI law firm Crosby buys malpractice cover as risk migrates to cyber insuranceNov 4, 2026 · Money
- Democrats promise AI regulation as they prepare to take control of US CongressNov 4, 2026 · Jog
- Democratic midterm win hits tech stocks on fears of AI regulationNov 3, 2026 · Agent Age
- California enacts new AI laws on workplace practicesNov 2, 2026 · Jog
- Hungarian legal articles: company executives are personally liable for AI decisionsNov 2, 2026 · Jog
- What does an AI lawyer have to do with bookkeeping? A Hungarian trade articleNov 1, 2026 · Money
- 115 new state AI laws in 30 states, with chatbot safety at the centreOct 26, 2026 · Jog
Also known as
AI regulation · AI regulation · AI regulation · regulation of artificial intelligence · AI regulation · regulation of artificial intelligence · AI regulations · AI law
Sources
This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.