AI regulation
In plain wordsLaws and binding government rules about artificial intelligence: what AI may be used for, what checks its makers must run, and what penalties apply when they break the rules.
Legally binding rules made by public authorities that govern the development, placing on the market or use of AI systemsAI system, n.: A complete piece of technology that uses AI to do something, such as screening loan applications, driving a car or chatting with customers, including the AI model and the software around it. and are backed by enforcement and sanctions. Regimes differ along several axes: horizontal laws spanning all sectors (the EU AI ActEU AI Act, prop. n.: The European Union's law on artificial intelligence, the first broad AI law anywhere: it bans some uses, sets strict checks for risky ones and puts duties on the makers of the biggest AI models., South Korea's AI Basic Act) versus sector-specific rules for medicine, credit or hiring; risk-basedrisk-based approach, n.: A way of regulating AI in which the strictness of the rules depends on how much harm a use could cause: bans for the worst uses, heavy checks for risky ones, little or nothing for the rest. tiers keyed to uses versus thresholds keyed to models, such as compute thresholdscompute threshold, n.: A legal cutoff based on how much computation was used to train an AI model. Models trained with more than that amount face extra rules, on the theory that bigger training runs produce more capable models. for frontier AIfrontier AI, n.: The most capable general-purpose AI systems of the moment — those at or beyond the leading edge — which governments single out for extra safety attention.; ex ante duties such as conformity assessmentconformity assessment, n.: The check, done by the maker or by an approved outside body, that a high-risk AI system meets the EU's legal requirements before it can be sold or used. versus ex post liabilityAI liability, n.: Who has to pay or answer in court when an AI system causes harm — the company that built it, the business that used it or the person who relied on it — and under which laws.; and disclosure duties versus outright bans (prohibited AI practicesprohibited AI practices, n.: Uses of AI that the EU bans outright as too harmful — such as manipulating people, social scoring, and most live facial recognition by police in public places.). Much AI regulation is also existing law — data protection, consumer protection, anti-discrimination, copyright — applied to AI.
AI regulation in the EU is horizontal, whereas the US relies mainly on sector regulators and state laws.
Origin
Legal scholarship on how to regulate AI took off in the mid-2010s; Matthew Scherer's 'Regulating Artificial Intelligence Systems' (Harvard Journal of Law & Technology, 2016) was an influential early analysis. China adopted binding rules on recommendation algorithms (2022), deep synthesis (2023) and generative AI (2023); the first comprehensive horizontal statute was the EU AI Act, proposed in April 2021 and adopted in 2024.
Usage note
Narrower than AI governanceAI governance, n.: The rules, institutions and habits that shape how AI is built and used — from laws and treaties to technical standards and a company's own review processes., which also covers standards, voluntary commitments and corporate practice. 'Regulating AI' may target models, applications, uses or the organizations behind them; much of the policy debate is about which target is right.
Related
- AI governanceAI governance, n.: The rules, institutions and habits that shape how AI is built and used — from laws and treaties to technical standards and a company's own review processes.
- EU AI ActEU AI Act, prop. n.: The European Union's law on artificial intelligence, the first broad AI law anywhere: it bans some uses, sets strict checks for risky ones and puts duties on the makers of the biggest AI models.
- risk-based approachrisk-based approach, n.: A way of regulating AI in which the strictness of the rules depends on how much harm a use could cause: bans for the worst uses, heavy checks for risky ones, little or nothing for the rest.
- compute thresholdcompute threshold, n.: A legal cutoff based on how much computation was used to train an AI model. Models trained with more than that amount face extra rules, on the theory that bigger training runs produce more capable models.
- AI liabilityAI liability, n.: Who has to pay or answer in court when an AI system causes harm — the company that built it, the business that used it or the person who relied on it — and under which laws.
- AI preemptionAI preemption, n.: The push in the United States for a single national AI law that would override the states' own AI laws, and the fight over whether that is wise or even lawful.
- Brussels effectBrussels effect, n.: The way European Union rules end up shaping products and laws worldwide, because global companies often find it simpler to follow the strictest big market's rules everywhere.
- Collingridge dilemmaCollingridge dilemma, n.: The trap that a new technology is easiest to control when little is known about its harms, but by the time the harms are clear it is too widespread to change easily.
Sources
- Scherer, M. U. (2016). Regulating Artificial Intelligence Systems: Risks, Challenges, Competencies, and Strategies. Harvard Journal of Law & Technology, 29(2), 353–400.
Mentioned in
AI governanceAI governance, n.: The rules, institutions and habits that shape how AI is built and used — from laws and treaties to technical standards and a company's own review processes., AI preemptionAI preemption, n.: The push in the United States for a single national AI law that would override the states' own AI laws, and the fight over whether that is wise or even lawful., Brussels effectBrussels effect, n.: The way European Union rules end up shaping products and laws worldwide, because global companies often find it simpler to follow the strictest big market's rules everywhere., ethics washingethics washing, n.: Using talk of ethics, such as principles, advisory boards or pledges, to look responsible and head off regulation, without actually changing what a company does., risk-based approachrisk-based approach, n.: A way of regulating AI in which the strictness of the rules depends on how much harm a use could cause: bans for the worst uses, heavy checks for risky ones, little or nothing for the rest., AI Basic ActAI Basic Act, prop. n.: South Korea's main AI law, in force since January 2026, which sets safety and transparency duties for AI used in sensitive areas and for AI-generated content., Collingridge dilemmaCollingridge dilemma, n.: The trap that a new technology is easiest to control when little is known about its harms, but by the time the harms are clear it is too widespread to change easily.