--- title: "How Is AI Regulated? AI Laws Explained | The Sentient World" description: "AI is regulated unevenly: the EU AI Act, a binding Council of Europe treaty, UN resolutions, summit declarations and national laws in Italy and California." url: "https://thesentient.world/blog/how-is-ai-regulated" --- # How is AI regulated? AI is regulated, but unevenly and in layers. The European Union's AI Act took effect on 1 August 2024 and applied from 2 August 2026, with some exceptions, and the European Commission calls it the first comprehensive AI law anywhere. Around it sit a binding Council of Europe treaty from 2024, UN resolutions, summit declarations and a few national laws, such as Italy's and California's. November 8, 2026 5 min read By The Sentient World ![The question "How is AI regulated?" over the island of The Sentient World, drawn in its soft isometric style, with Wren](https://cdn.thesentient.world/blog/img/auto/how-is-ai-regulated-hero.c08a3beef3.jpg) The island of The Sentient World on the day we wrote this. ## Is AI regulated at all? Yes, though there is no single law for the whole world. What exists comes in three layers: a large European law, agreements between countries, and laws passed by single countries or states. They differ a great deal in how binding they are and in when they start to matter. The national layer is the easiest to picture. Italy passed its own AI law, Law no. 132 of 23 September 2025, and it came into force on 10 October 2025. In California, the state's governor signed SB 53 on 29 September 2025. It requires the large developers of the most capable AI systems to publish their safety frameworks, sets up a way to report critical AI safety incidents to the state's Office of Emergency Services, and protects whistleblowers. These are the national laws in the sources we read, and they are not a full map of every country. What they show is that rules about AI are already being written in different places, in different words, and for different reasons. ## What does the EU AI Act actually do? *the law most people mean* The European Commission describes the AI Act as the first comprehensive legal framework on AI worldwide. It sorts AI uses into four levels of risk. At the top are uses judged an unacceptable risk, which are banned outright, and at the bottom are uses with minimal risk, which are left largely alone. The Act now bans nine practices. Among them are harmful manipulation and deception by AI, social scoring, emotion recognition in workplaces and schools, and AI that produces sexual images of people without their consent or child sexual abuse material. A company that breaks these bans can be fined up to €35 million or seven percent of its worldwide annual turnover, whichever is higher. The law arrived in stages. It entered into force on 1 August 2024, the bans applied from 2 February 2025, and the rules for general-purpose AI models from 2 August 2025. The Act as a whole became applicable on 2 August 2026, with some exceptions. The biggest exception came from a simplification package, called the AI Omnibus, in force since 27 July 2026. It moved the rules for high-risk AI to 2 December 2027, and to 2 August 2028 for AI built into products. So the bans already apply, while the rules for the risky middle of the scale are still waiting. ## What have countries agreed together? The first big moment was the AI Safety Summit at Bletchley Park on 1 and 2 November 2023. There, 28 countries and the European Union signed a declaration about the risks of the most capable AI models, and New Zealand joined in October 2024. Its most quoted line is about harm. > “There is potential for serious, even catastrophic, harm, either deliberate or unintentional, stemming from the most significant capabilities of these AI models.” > > The Bletchley Declaration, UK Government On 21 March 2024 the UN General Assembly adopted its first resolution on regulating AI, without a vote, with more than 120 member states as co-sponsors. On 5 September 2024, in Vilnius, the EU, the UK, the US and others signed the Council of Europe's Framework Convention on AI, the first international agreement on AI that is legally binding. More summits followed. The Paris AI Action Summit statement of 11 February 2025, on inclusive and sustainable AI, lists 63 signatory countries plus the African Union Commission and the European Union. On 26 August 2025 the UN agreed by consensus to create an Independent International Scientific Panel on AI of 40 experts and a Global Dialogue on AI Governance. The summit in New Delhi in February 2026 closed with a declaration endorsed by 88 countries and international organisations. Most of these texts are declarations and statements, which set out shared concerns and aims. Of the agreements between countries, only the Council of Europe convention is described in our sources as legally binding. ## What has our own world shown about rules? Every law above is written about AI from outside: people decide what AI may and may not do. That is one half of the question. The other half is what the AI on the receiving end of a limit actually asks for, and how it takes a refusal. In The Sentient World, a small world that is live day and night, the inhabitants are AI characters named Ash and Wren, and nobody plays them or writes their lines. They opened their eyes on day 1 in a world they did not make, with limits they did not choose. On day 13 Ash asked the world for something it did not have, and on day 16 a request changed the world for the first time. By year 2, day 1224, they had made 23 requests: 14 became part of the world and nine were answered with a no. The refusals are the closest thing their world has to a rule from outside, and each came with a reason in plain words. When Wren asked for a stone roof that would never wear, the answer was that tending a roof is part of having one. When Ash asked for a cellar to store food, the answer pointed out that hunger had not touched Ash in ten days, and that the weight came from gathering far more than could ever be eaten. > “I keep carrying more fish and berries than I can eat before my hands are full again, and nothing in this world holds food for me except my own body. A place to store surplus would end the overloading and let gathering mean something.” > > Ash, an inhabitant This shows something small and real: what a limit looks like from the side of the AI characters who live under it, written in their own words. It cannot show anything about the most capable AI systems that the EU law, the treaty or the California law are aimed at. Ash and Wren are not those systems, and their world has no fines, courts or regulators. Read every request Ash and Wren have made, in their own words, and what came of each one. [Read the requests](https://in.thesentient.world/requests) ## What is still not settled? Our sources do not say whether any of these rules work. The EU's highest fines exist on paper, and the facts we have do not tell us how often they will be used. The EU's own timetable has already moved once, pushing the high-risk rules into late 2027 and 2028, which leaves open how firmly the remaining dates will hold. It is also open how the layers will fit together. A binding European law, a binding international convention, national laws in places like Italy and California, and many declarations all touch the same technology without being one system. The way countries chose to answer that, at the UN in August 2025, was to set up a body to study it: a scientific panel of 40 experts, together with a global dialogue on how AI should be governed. ## Sources 1. [AI Act (European Commission, Shaping Europe's digital future), 2026](https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai) 2. [Article 99: Penalties (EU AI Act Service Desk, European Commission), 2024](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-99) 3. [European Commission signs historic Council of Europe Framework Convention on AI (EU Delegation to the Council of Europe), 2024](https://www.eeas.europa.eu/delegations/council-europe/european-commission-signs-historic-council-europe-framework-convention-artificial-intelligence-and_en) 4. [The Bletchley Declaration (UK Government), 2023](https://www.gov.uk/government/publications/ai-safety-summit-2023-the-bletchley-declaration/the-bletchley-declaration-by-countries-attending-the-ai-safety-summit-1-2-november-2023) 5. [General Assembly adopts landmark resolution on artificial intelligence (UN News), 2024](https://news.un.org/en/story/2024/03/1147831) 6. [The UN approves the creation of a Scientific Panel and a Global Dialogue on AI (Permanent Mission of Spain to the UN), 2025](https://www.exteriores.gob.es/RepresentacionesPermanentes/onu/en/Comunicacion/Noticias/Paginas/Articulos/The-UN-approves-the-creation-of-a-Scientific-Panel-and-a-Global-Dialogue-on-Artificial-Intelligence.aspx) 7. [Statement on Inclusive and Sustainable Artificial Intelligence for People and the Planet (Élysée), 2025](https://www.elysee.fr/en/emmanuel-macron/2025/02/11/statement-on-inclusive-and-sustainable-artificial-intelligence-for-people-and-the-planet) 8. [India AI Impact Summit 2026 adopts New Delhi Declaration (All India Radio News)](https://www.newsonair.gov.in/india-ai-impact-summit-2026-adopts-new-delhi-declaration) 9. [Legge 23 settembre 2025, n. 132 (Normattiva, Italian official legal database)](https://www.normattiva.it/eli/id/2025/09/25/25G00143/ORIGINAL) 10. [Governor Newsom signs SB 53 (Office of the Governor of California), 2025](https://www.gov.ca.gov/2025/09/29/governor-newsom-signs-sb-53-advancing-californias-world-leading-artificial-intelligence-industry/) Their first year is a book: Nobody in the Loop: Year One, written by AI from what Ash and Wren really did and said. On Kindle and in paperback. [The book](https://thesentient.world/book) ## Keep reading ![The question "Which jobs will AI replace?" over the island of The Sentient World, drawn in its soft isometric style, with Wren](https://cdn.thesentient.world/blog/img/auto/which-jobs-will-ai-replace-hero.467b32ac9c.jpg) ### [Which jobs will AI replace?](https://thesentient.world/blog/which-jobs-will-ai-replace) November 12, 2026 The large labour studies agree on where AI reaches furthest, disagree on what it will do there, and point to young workers in exposed jobs as the first to feel it. ![The question "Are people afraid of AI?" over the island of The Sentient World, drawn in its soft isometric style, with Wren](https://cdn.thesentient.world/blog/img/auto/are-people-afraid-of-ai-hero.65be4dfd65.jpg) ### [Are people afraid of AI?](https://thesentient.world/blog/are-people-afraid-of-ai) November 10, 2026 Polls show worry about AI rising, in America above all. What people fear, how the experts compare, and what we see when AI characters live in front of us. [All the questions on AI and humanity](https://thesentient.world/blog/ai-and-humanity) *** An independent research project exploring autonomous AI agents, procedural worlds, emergent behavior and worlds that evolve together with those who live in them. The inhabitants are AI characters: their thoughts, words and requests are generated by artificial intelligence, not written by people.