# AI regulation

> The laws that set what AI systems may do, and who answers when they go wrong.

- Level: Basic
- Topics: AI
- URL: https://tidbit.legios.com.ar/ai-regulation/
- In Spanish: [Regulación de la IA](https://tidbit.legios.com.ar/es/ai-regulation/)

**AI regulation** is the set of laws and rules governments write for artificial intelligence: what is forbidden, what needs extra checks, what people must be told. It is not one law but a patchwork that is changing quickly, because AI now helps decide who gets a loan, a job interview or a medical test.

The most complete example is the European Union's **AI Act**, which entered into force in August 2024 and whose obligations kick in gradually over several years. It sorts uses by risk. A few are banned outright, such as social scoring by governments or systems that manipulate people by exploiting their vulnerabilities. High-risk uses, like screening CVs, grading exams or deciding on credit, must meet requirements on data quality, human oversight and documentation. Chatbots and deepfakes carry transparency duties: people must know they are talking to a machine or looking at AI-made content. Everyday uses like spam filters get no new obligations. Fines for the worst violations can reach 35 million euros or 7% of a company's worldwide annual turnover.

Elsewhere the picture is mixed. The United States has no single federal AI law and relies on existing rules plus a growing number of state laws; China has specific rules for generative AI and recommendation algorithms; Brazil, Canada and others have debated their own bills. Older laws also still apply to AI: privacy rules like GDPR, consumer protection, anti-discrimination. For a company, the practical question is less "is AI legal?" and more "what does this particular use require?"

## An analogy

Food safety rules. Nobody bans cooking, but a restaurant gets inspected, a baby formula maker gets inspected far more strictly, and a few ingredients are simply forbidden. The AI Act applies the same idea by level of risk.

## Keep in mind

- The EU AI Act sorts AI uses by risk: banned, high, limited, minimal.
- High-risk uses (hiring, credit, exams) need human oversight and documentation.
- People must be told when they talk to a bot or see AI-made content.
- Privacy, consumer and anti-discrimination laws already apply to AI.

## Go deeper

- [AI Act (European Commission)](https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai)
- [Regulation of artificial intelligence on Wikipedia](https://en.wikipedia.org/wiki/Regulation_of_artificial_intelligence)
- [Artificial Intelligence Act on Wikipedia](https://en.wikipedia.org/wiki/Artificial_Intelligence_Act)

## Related terms

- [Artificial intelligence](https://tidbit.legios.com.ar/artificial-intelligence/): Getting computers to do things that, until recently, only people could do.
- [Vulnerability and CVE](https://tidbit.legios.com.ar/vulnerability-and-cve/): A flaw that lets an attacker misuse software, and the public id that names it.
- [Data quality](https://tidbit.legios.com.ar/data-quality/): How far data can be trusted for the job it is used for.
- [Chatbot](https://tidbit.legios.com.ar/chatbot/): A program you talk to by typing or speaking, as if it were a person.
- [Deepfake](https://tidbit.legios.com.ar/deepfake/): A fake video, photo or voice made with AI to look and sound like a real person.
- [Spam](https://tidbit.legios.com.ar/spam/): Unwanted messages sent in bulk: ads, scams and junk you never asked for.
- [Generative AI](https://tidbit.legios.com.ar/generative-ai/): AI that creates new text, images, audio or code instead of just sorting things.
- [Recommendation system](https://tidbit.legios.com.ar/recommendation-system/): Software that guesses what you will want next: a song, a show, a product.