The laws and rules governments set for how AI is built and used.
AI regulation means the laws and rules a country or region sets so that AI doesn't harm people. The EU's AI Act entered into force in August 2024. It sorts uses of AI by risk (banned, high-risk, transparency duties and minimal risk), sets stricter rules the riskier a use is, and phases its obligations in step by step.
South Korea passed its AI Basic Act in December 2024, and it took effect on January 22, 2026. It requires people to be told when content was made with generative AI, and puts extra duties on "high-impact" AI used in areas such as healthcare and hiring. China has applied rules to generative AI services since August 2023.
Think of it like this
Think of age ratings for films and games. The more intense the content, the stricter the rules about who can see it, and some things aren't allowed at all. The EU's AI Act works in a similar way, with tighter rules for riskier uses of AI.
For example
The EU AI Act bans some uses of AI outright, such as social scoring that ranks people in ways that can lead to unfair treatment.
Common mix-ups
Mix-upAI regulation puts every AI under the same rules.
ActuallyKorea's AI Basic Act also puts its extra safety duties only on very large models, whose training took at least 10^26 calculations. Even within one law, what you must do depends on an AI's size and use.
Mix-upKorea doesn't have an AI law yet.
ActuallyThe AI Basic Act is already in force, but investigations and fines come with a guidance period of at least a year, so saying "punishment starts straight away in 2026" isn't accurate. It also requires foreign AI companies above certain thresholds to appoint a representative in Korea.
Try it in class
For teachers at the start of a lesson — or for students working on their own.
- Question
- If you rated the AI used at your school like film age ratings, where would each one go?
- Activity
- Draw four boxes on the board — prohibited, high-risk, transparency obligations, minimal risk. Each group writes five AI uses from school and daily life on cards (lunch suggestions, face recognition for attendance, a homework helper…) and sticks them in a box. Explain your choices, then mark the start dates from the EU (E0413), Korea (E0830) and China (E0243) records on one timeline.
Records about this
Source-checked moments from AI history.
However, fact-finding investigations and fines come with a guidance period of at least 1 year, so statements that penalties begin in 2026 are inaccurate
Entry into force is not application. Application is phased
The world's first regulation dedicated to generative AI