Taiwan's AI Basic Act: What Its 20 Articles Do and Do Not Say

Taiwan's 20-article AI Basic Act sets seven principles and agency duties but no fines. A plain reading of the official text, with its limits.

Guide: AI BasicsLast verified: October 20266 min read

Taiwan's Artificial Intelligence Basic Act is a 20-article framework law that the Legislative Yuan passed on December 23, 2025, and that took effect on January 14, 2026, the day it was promulgated. It names seven principles for AI development, makes the National Science and Technology Council the central authority, and orders the government to bring existing laws into line within two years. It does not set fines or tiers of risk itself, and it leaves most detail to later rules from the Ministry of Digital Affairs and sector regulators.

Key takeaways

  • The Act is mainly a mandate for government: most duties fall on agencies, not on companies.
  • Article 3 defines AI as a system that uses machine learning and algorithms to generate predictions, content, recommendations, or decisions.
  • The government has two years from the effective date, so until January 2028, to enact, amend, or repeal laws that conflict with the Act.
  • High-risk applications are to carry advisory notices or warnings, and the risk categories themselves are left to the Ministry of Digital Affairs and sector authorities.
  • The official database carries an English text, and it states that the Chinese version prevails if the two differ.

What the law is, and why the name "AI Tai" leads to it

No product, company, or technical standard called "AI Tai" could be verified in any primary source. The closest topic that can be documented is Taiwan's AI Basic Act, which shows how one jurisdiction has written AI policy into statute. The Act was enacted on the Executive Yuan's draft, which the cabinet approved on August 28, 2025, according to a Taiwanese legal research institute. After the third reading on December 23, 2025, the president promulgated it on January 14, 2026. Article 20 makes it effective on promulgation.

This Act sets direction and assigns responsibilities. It is not a code of conduct for businesses. Article 1 says matters the Act does not cover remain governed by other laws, so privacy, consumer, labor, and sector statutes keep applying to AI uses as before.

The seven principles in Article 4

Article 4 tells the government what to follow while it promotes AI research and use. The principles are sustainable development and well-being, human autonomy, privacy protection and data governance, cybersecurity and safety, transparency and explainability, fairness and non-discrimination, and accountability. Some carry concrete content. Human autonomy includes ensuring human oversight. The data principle adopts data minimization and also encourages reuse of non-sensitive data. Transparency asks that AI outputs come with appropriate disclosures or labels so that people can assess risk.

These are directions for policy. A person or company cannot be fined for failing to be "fair" under Article 4 alone. The principles matter because agencies must follow them when they write the detailed rules that do bind private parties.

Who does what: a map of the articles

The table summarizes the operative articles from the official English text. It is a simplified reading, and the Chinese text controls.

ArticleSubjectWhat it provides
2AuthoritiesThe National Science and Technology Council is the central competent authority; local governments act locally; other central agencies handle matters in their own fields
5Harm and high-risk noticesGovernment must ensure AI use does not infringe life, body, liberty, or property, or involve bias, false advertising, or misleading information; high-risk products carry advisory notices or warnings; children and adolescents' best interests come first
6Strategy committeeThe Executive Yuan sets up a National AI Strategic Committee, convened by the Premier, meeting at least once a year
11SandboxesSector authorities may create safe experimentation environments for innovative AI products
14Personal dataAvoid unnecessary collection and use; promote privacy by design and by default
15LaborUse AI to safeguard labor rights; bridge skills gaps; give employment counseling to people who lose work because of AI
16Risk frameworkThe Ministry of Digital Affairs promotes a risk taxonomy interoperable with international frameworks; sector authorities write risk-based rules
17LiabilityFor high-risk AI, government must clarify who is liable and set up relief, compensation, or insurance mechanisms
18Legal reviewExisting laws and measures reviewed and conformed within two years of the effective date
19Government useAgencies using AI assess risks, plan responses, and set usage guidelines or internal controls

What the Act leaves open

Three gaps stand out. First, no article contains a fine, a criminal penalty, or an enforcement agency with inspection powers, so the Act does not by itself punish any conduct. Article 16 says the circumstances listed in Article 5, paragraph 1, are to be restricted or prohibited "in accordance with the law", which points to other statutes and future rules. Second, the Act does not define "high-risk" in its text. Article 5 assigns that designation to the central authority for the relevant industry, in consultation with the Ministry of Digital Affairs. Third, Article 17 promises clarity on liability for high-risk AI but does not say who pays or how. It also exempts research and development before actual use, unless the work is tested in a real-world environment or its results are used to provide products or services.

Legal commentary published after passage expected sector rules in finance, health, employment, and consumer protection within the two-year window. That expectation is analysis, not statute. Whether and when the Ministry of Digital Affairs published its risk framework could not be confirmed from an official source for this article.

Reading a framework law like this one

  • Check whether duties are placed on agencies or on private parties.
  • Look for penalties, enforcement bodies, and deadlines in the text itself.
  • Note which terms, such as "high-risk", are defined in the law and which are delegated.
  • Find the effective date and any review clauses, such as the two-year clock in Article 18.
  • Confirm which language version is authoritative, and whether the translation is official.

This article gives general information about a foreign statute and is not legal advice. A business with customers or operations in Taiwan should rely on the Chinese text, current agency rules, and qualified local counsel.

FAQ

Is "AI Tai" a product or a company?

No such product, company, or standard could be verified. The nearest documented subject is Taiwan's AI Basic Act, which is what this page explains.

Does the Act ban any AI uses?

Not directly. Article 5 lists outcomes the government must prevent, such as infringement of life, body, liberty, or property, and Article 16 says those circumstances are to be restricted or prohibited in accordance with the law. The bans themselves would have to come from other laws or later rules.

How does the Act define artificial intelligence?

Article 3 defines it as a system capable of autonomous operation that, through input or sensing and by means of machine learning and algorithms, generates outputs such as predictions, content, recommendations, or decisions that affect physical or virtual environments.

Does it apply to companies outside Taiwan?

The text does not set rules for foreign companies. Obligations that reach private parties are expected to come from sector rules, so their reach will depend on those rules and on the usual jurisdiction of Taiwanese law.

Last verified: October 2026

Sources

New to artificial intelligence?

Start with the basics, then follow the guides to tools, rules, and effects on society.

Start with AI Basics