Industrial AI Rules in 2026: The EU AI Act Delay and Korea's AI Basic Act
Industry Operations Note special edition: the industrial AI regulation timeline
The EU AI Act's high-risk obligations were due to apply on August 2, 2026. They now apply from December 2, 2027 for standalone high-risk AI and from August 2, 2028 for high-risk AI built into products. Korea's AI Basic Act has been in force since January 22, 2026, and the government has set a grace period of at least one year before enforcement such as administrative fines.
This article was written against the amending regulation published in the EU Official Journal and materials from Korea's Ministry of Government Legislation and Ministry of Science and ICT, as of September 19, 2026. It is not legal advice. Whether your systems are covered should be decided from the source texts and professional review.

Illustrative image. It does not depict any actual authority, regulatory document, or specific workplace.
EU: high-risk obligations pushed back in two steps
The change started with the European Commission's Digital Omnibus proposal on AI. Delays in designating national supervisory authorities and preparing harmonised standards are cited as the reason.
The European Parliament gave final approval on June 16, 2026, and the Council on June 29. The text was published in the Official Journal as Regulation (EU) 2026/1744 on July 24 and entered into force on July 27, six days before the original August 2 date for high-risk obligations.
The new dates split in two. Standalone high-risk AI listed in Annex III, such as uses in hiring, education, and access to essential services, must meet the requirements from December 2, 2027. High-risk AI in Annex I, used as a safety component of products covered by EU product safety law, follows from August 2, 2028.
AI in machinery moves under the Machinery Regulation
This is the most direct change for manufacturers putting AI into equipment. The amendment moves Machinery Regulation (EU) 2023/1230 from Section A to Section B of AI Act Annex I. High-risk AI in machinery no longer takes the AI Act's technical requirements directly. Instead, the Commission will fold AI requirements into the Machinery Regulation through delegated acts by August 2, 2028.
The requirements did not disappear; their legal home changed. Some commentary also says the definition of a safety component was narrowed to cases intended to prevent or mitigate health and safety risks, so check the final text to see whether your AI falls in scope. The Machinery Regulation has its own application date, which is worth confirming at the same time.
What was delayed and what was not
Commentators widely describe the amendment as a delay, not an exemption. Rules already in application, such as prohibited AI practices and obligations for general-purpose AI models, are described as outside the delay.
Some burdens eased. Commentary reports that simplified technical documentation and proportionate quality management now extend to small mid-cap companies slightly larger than SMEs, and that the AI literacy provision was softened from ensuring to supporting. Check the thresholds against the source text.
Korea: the AI Basic Act is already in force
The Framework Act on the Development of Artificial Intelligence and the Establishment of a Foundation for Trust (Act No. 20676) was promulgated on January 21, 2025 and took effect on January 22, 2026. Its Enforcement Decree is reported to have taken effect the same day.
The Act defines high-impact AI as AI that may significantly affect, or pose risks to, human life, physical safety, or fundamental rights, and lists 11 areas: energy supply, drinking-water production, healthcare, medical devices, nuclear safety management, biometrics for criminal investigation, hiring, loan screening, and similar decisions, the core operation of transport means, facilities, and systems, public decision-making, student assessment, and areas set by Presidential Decree.
General manufacturing equipment or process monitoring is not named as a separate item on that list. The same AI may be treated differently when used in energy supply or transport facility operation, and coverage has to be judged through the Ministry of Science and ICT's guidelines and confirmation procedure. Do not conclude that you are out of scope just because manufacturing is not listed.
Operator duties and the grace period
The baseline duties are advance notice that high-impact or generative AI is in use, and labeling of generated output. High-impact AI adds a risk management plan, an explanation plan, user protection, human oversight, and documentation, and the Act includes administrative fines for breaches such as the notice duty.
The Ministry of Science and ICT said it will hold off on enforcement for at least one year, with fact-finding investigations limited to exceptional cases such as loss of life or human rights violations. No end date for the grace period has been announced, so treat the current period as preparation time.
AI regulation and occupational safety regulation rest on different laws, so review the two sets of duties separately. Changes on the safety side are covered in the field guide 2026 Korean Workplace Safety Law Changes: Dates and Checklist.
Regulatory timeline, 2026-2028
| Date | Region | What happens |
|---|---|---|
| January 22, 2026 | Korea | AI Basic Act takes effect; enforcement grace period (at least one year) begins |
| July 27, 2026 | EU | Omnibus amendment, Regulation (EU) 2026/1744, enters into force |
| August 2, 2026 | EU | Original date for high-risk obligations (postponed) |
| December 2, 2027 | EU | Annex III standalone high-risk AI requirements apply |
| August 2, 2028 | EU | Annex I product-embedded high-risk AI requirements apply; deadline for Machinery Regulation AI delegated acts |
Source: Ministry of Government Legislation (promulgated January 21, 2025), Ministry of Science and ICT (January 2026), EU Official Journal, Regulation (EU) 2026/1744 (July 24, 2026)
What to do now
- List the AI running in your equipment and processes, and classify each by use.
- For products exported to the EU, separate AI used in machinery safety functions from standalone uses.
- Check whether domestic uses overlap high-impact areas such as energy supply or transport facility operation, and consider the government confirmation procedure when unsure.
- Set up records that keep AI outputs together with the actions people took.
- Decide who reviews AI alerts, who makes the final call, and who is responsible for oversight.
- If you use generative AI for customer support or documents, review how you give notice and label output.
- Assign someone to track the Machinery Regulation delegated acts and harmonised standards.
Summary
The EU has moved high-risk AI obligations to December 2027 and August 2028, and Korea's AI Basic Act is in force with a grace period. The delay and the grace period are both preparation time, so use them to put AI use classification, record keeping, and human oversight in place first. The source texts are available from Korea's National Law Information Center and EUR-Lex.
Note 6 of 6, Industrial Monitoring Pilots: How to Move From PoC to Operations, covers who reviews alerts and how they connect to inspections.
Official sources
- EUR-Lex, Regulation (EU) 2026/1744
- EUR-Lex, consolidated AI Act (as of July 27, 2026)
- Council of the EU, final approval press release (June 29, 2026)
- National Law Information Center, Framework Act on AI
- Korea policy briefing, Ministry of Science and ICT announcement on the AI Basic Act
- Future of Privacy Forum, AI Act timeline under the AI Omnibus
- Orrick, Digital Omnibus final compliance changes