Explain like I'm five · AI Theory

What is the AI Basic Act and High-Impact AI?

Korea's AI Basic Act, told as a story about a legal floor every inspection station must clear — and the heavier duties high-impact AI carries on top.

our station does it our own wayso do we, our own wayus too, our own waythere was a time when every station just made up its own rulescustomers had no way to know which station was actually safe

There was a time when every station just made up its own rules.Customers had no way to know which station was actually safe.

I took the loss — whose fault is it, exactly?no AI-specific law to point to — no answer from mewith no AI-specific law, it was hard to say after an accident who should have done what

With no AI-specific law, it was hard to say after an AI accident who should have done what.Even after an accident, no shared legal standard said what an AI business should have done.

AI BASIC ACT — THE FLOOR EVERY STATION MUST CLEARproviders of generative or high-impact AI give notice firstlisted area + big stake in life, safety, rights = high-impacthigh-impact AI carries extra duties on topthe AI Basic Act is the legal floor every station has to clearin areas the law lists, an AI that heavily affects life, safety, or rights is high-impact, with heavier duties

The AI Basic Act is the legal floor every station has to clear.In areas the law lists (medicine, hiring, loan screening, and others), an AI that heavily affects life, safety, or basic rights is classified as high-impact AI, with heavier duties.

Give advance noticetelling people up front that AI is making the call
Set up a risk-management planhigh-impact AI providers must set up and run one
Put safety measures in placeactually putting safeguards in place
Keep human oversightpeople watch the AI and can step in
ORDINARY AI vs HIGH-IMPACT AIother AI — no added duties like high-impact AI hashigh-impact AI — in listed areas, when life, safety, or rights are heavily at stakehigh-impact AI — risk plan, explanation, user protection, human oversight on tope.g. AI for medical devices, hiring, loan screening, or biometric crime probes

Not every AI carries the same weight of duty.The more an AI affects people, the more safeguards the law demands.

the law draws the floorwhat goes above the floor is each station's own jobthe law is only the floor — real safety still comes from ethics standards and risk management

Obeying the law doesn't automatically make an AI safe.The grading principles (ethics standards) and operations map (risk management) above the legal floor are still each station's own responsibility.

In one breath

The AI Basic Act = the legal floor every station must clear, with high-impact AI carrying heavier duties on top.

Korea's AI Basic Act covers both promoting AI and building trust in it. It requires notice when high-impact or generative AI is used, and classifies AI in listed domains — energy, drinking water, medicine, nuclear facilities, biometrics in criminal investigation, hiring and loan screening, transport, public services, student evaluation, and others — as 'high-impact AI' where life, bodily safety, or basic rights are heavily at stake. It requires measures like a risk-management plan, an explanation plan, user protection, human oversight, and record-keeping, and asks businesses to make an effort at a fundamental-rights impact assessment. It took effect on January 22, 2026. The law only draws the floor; the detailed principles above it (AI ethics and governance) and the execution procedure (an AI risk management framework) are still something an organization has to build for itself.

When grown-ups say it

AI Basic Act
The minimum grading rule the country itself sets. A common law for AI businesses, though the weight of duty differs by type of AI.
High-impact AI
AI that heavily affects life, safety, or basic rights. Carries extra duties on top of the basics.
Advance notice duty
Telling people up front that AI is making the call. Lets the customer know and choose with open eyes.
Risk-management duty
The duty of high-impact AI providers to set up and run a risk-management plan. Similar work to the stages of the operations map. → The Operations Map That Walks the Whole Station
Safety assurance measures
Actual safeguards put in place to prevent accidents. Principle alone isn't enough — execution has to follow.
Human oversight
The mechanism that lets people watch the AI and step in when needed. Shares its roots with the human review the ethics charter guarantees. → The Grading Charter Every Station Shares
Legal floor
The minimum level the law guarantees. What's built above it is up to each organization's own ethics and risk management.
Administrative Fines & Corrective Orders
The legal consequence of failing these duties. Breaches such as the notice duty can draw fines of up to 30 million won; the government committed to holding off on fines for at least a year after the Act took effect.
Next story
The Eye That Used to Just Look Grows Arms and Legs →
Want another term as a picture book? Request →