Skip to content
EN

Lesson 1 of 8 · 6 min · +100 XP

What is the EU AI Act, in short?

In short, the EU AI Act sorts AI by risk. Unacceptable uses, such as social scoring, are banned. High-risk uses, such as CV screening or credit scoring, need testing, documentation and human oversight. Chatbots and deepfakes must be disclosed. Most AI, like spam filters, has no extra rules. The UK and US mainly rely on existing laws and regulators.

▶ Try it

Why it matters

AI rules are being written right now, and they affect what apps can do with you. The EU AI Act is the biggest worked example: it applies to anyone offering AI in the EU, including US and UK companies. Knowing the risk levels helps you ask the right questions.

In 6 minutes you'll be able to

  • EU AI Act levels: banned, high risk, must disclose, minimal risk.
  • AI that decides on jobs, loans or exams has the strictest rules.
  • The UK and US mostly use existing laws and regulators, so those rights still apply.

What you'll practise

  1. Four risk levels

    The EU AI Act puts every AI use on a ladder. The higher the step, the stricter the rules.

    🚫 Unacceptable: banned
    AI that scores people's social behaviour, manipulates people in harmful ways or reads emotions at work or school. Banned in the EU since 2 February 2025.
    ⚠️ High risk: strict rules
    AI that helps decide on jobs, loans, exams or access to services. It needs risk checks, good data, records and human oversight.
    💬 Transparency: tell people
    Chatbots must say they are AI, and deepfakes must be labelled as made or changed by AI.
    ✅ Minimal risk: no new rules
    Spam filters, game opponents, photo filters. Most AI in use today sits here.
    🌍 Who has to follow it
    Anyone who offers or uses AI in the EU, even a company based in the US or the UK.

  2. Which risk level?

    Sort each AI use into its EU AI Act risk level.

    • A city gives residents a "good citizen" score that affects their benefits → Banned
    • Software that reads employees' emotions from their webcam → Banned
    • A tool that ranks job applicants' CVs → High risk
    • A model that decides who gets a loan → High risk
    • AI that marks school exams → High risk
    • Lumi, a customer-service chatbot → Must disclose
    • An AI-made video of a real politician → Must disclose
    • A spam filter in your email → Minimal risk

  3. The EU timeline

    The law switches on in stages. Put the milestones in order, earliest first.

    1. The AI Act enters into force
    2. Bans on unacceptable-risk AI start
    3. Rules for general-purpose AI models apply
    4. Chatbot and deepfake transparency rules apply, and enforcement begins
    5. High-risk rules for uses like hiring, credit and education apply
    6. High-risk rules for AI built into products like toys and lifts apply

  4. UK, US, EU

    Different places, different approaches. What applies here?

    1. A US start-up sells its CV-screening AI to a company in France. Does the EU AI Act apply?

      • Yes, because the AI is used in the EU (Answer)
      • No, the company is American

    2. How does the UK mainly handle AI so far?

      • Existing regulators apply their rules to AI in their own areas (Answer)
      • It copied the EU AI Act word for word
      • AI is not covered by any UK law

    3. And in the US?

      • No single federal AI law, but existing laws and some state laws apply (Answer)
      • One federal AI law covers every use

Remember

Rules follow risk: the more an AI can affect your life, the more it has to prove.

Key facts

  • The EU AI Act entered into force on 1 August 2024, and its bans on unacceptable-risk AI have applied since 2 February 2025. Since a 2026 amendment, organisations must take measures to support their staff's AI literacy. · Source: Regulation (EU) 2024/1689 (AI Act), Article 4 as amended by Regulation (EU) 2026/1744
  • Rules for general-purpose AI models have applied since 2 August 2025. Transparency rules, such as disclosing chatbots and labelling deepfakes, apply from 2 August 2026, with a grace period to 2 December 2026 for machine-readable AI labels on systems already on the market. · Source: European Commission, AI Act application timeline
  • A 2026 amendment, the AI Omnibus, moved high-risk rules for uses like hiring, credit and education to 2 December 2027, and for AI built into products to 2 August 2028. · Source: European Commission, AI Omnibus enters into force, July 2026
  • The most serious breaches of the EU AI Act can be fined up to €35 million or 7% of a company's worldwide annual turnover, whichever is higher. · Source: Regulation (EU) 2024/1689 (AI Act), Article 99
  • The UK has no AI-specific act. Existing regulators, such as the data protection, financial and competition regulators, apply their rules to AI in their own sectors.

Updated:

Questions people ask

Does the EU AI Act apply to US and UK companies?

Yes, if they offer AI systems in the EU or their AI's output is used there. Where the company is based doesn't matter.

Does the UK have an AI law?

Not a single AI act. The UK relies on existing laws, such as data protection, equality and consumer law, enforced by sector regulators like the ICO, FCA and CMA.

Is there a federal AI law in the US?

No broad federal AI law has been passed. Existing laws on consumer protection, discrimination and credit apply to AI, and several states have passed their own AI rules.

When does the EU AI Act take effect?

In stages: bans from February 2025, general-purpose AI rules from August 2025, transparency rules and enforcement from August 2026, and high-risk rules from December 2027 or August 2028.

Related terms