EU AI Act After the Digital Omnibus: New Deadlines for Industrial AI

The Digital Omnibus on AI, Regulation (EU) 2026/1744, entered into force on 27 July 2026 and pushed back the EU AI Act's high-risk rules: stand-alone high-risk systems listed in Annex III now apply from 2 December 2027, and high-risk AI built into regulated products under Annex I from 2 August 2028. The bans, the (now softened) AI literacy duty and the general-purpose AI rules already apply, and the Article 50 transparency rules started on 2 August 2026.

Checked 1 October 2026. The dates below come from the European Commission's AI Act page and law-firm analyses of the published regulation, listed under Sources. This is an engineer's summary, not legal advice.

What did the Digital Omnibus change?

The Commission proposed the omnibus in November 2025 because the AI Act's support structures (national authorities, notified bodies and harmonised standards) were not ready for the original August 2026 high-risk deadline. Parliament and Council reached a political agreement on 7 May 2026, and the regulation was published in the Official Journal on 24 July 2026, entering into force three days later.

The changes that matter most to engineers:

• High-risk deadlines moved to fixed dates: 2 December 2027 for Annex III systems and 2 August 2028 for Annex I systems.
• Machinery left the AI Act's direct scope. The Machinery Regulation moved from Section A to Section B of Annex I (more on this below).
• AI literacy was softened. Providers and deployers must still take measures to support AI literacy among their staff, but the amended Article 4 states that it does not require them to guarantee any specific level for any individual.
• A new prohibition on AI systems that generate non-consensual intimate imagery or child sexual abuse material applies from 2 December 2026.
• Relief for smaller firms was extended to small mid-cap enterprises, with simpler documentation.
• Registration stays. Providers who decide an Annex III system is not high-risk under Article 6(3) must still register it in the EU database.
• Sandboxes: an EU-level regulatory sandbox run by the AI Office, and national sandboxes required by 2 August 2027.

What still applied from 2 August 2026?

The transparency rules in Article 50 were not delayed for new systems. From 2 August 2026:

• AI that interacts with people must be designed so they know they are dealing with an AI system, unless it is obvious. A maintenance assistant chatbot on the shop floor or a customer-support bot for your machines is covered.
• Deepfakes must be disclosed as artificially generated or manipulated.
• AI-generated content (text, images, audio, video) must be marked in a machine-readable way so it can be detected. Systems already on the market before 2 August 2026 have until 2 December 2026 for this Article 50(2) duty.

The AI Office's enforcement powers over general-purpose AI models and the governance bodies are also operating. For most industrial teams, Article 50 is the obligation that is live today.

Is my industrial AI high-risk?

There are two routes into high-risk, and most factory AI uses fall into neither.

Route 1, Annex I. An AI system is high-risk if it is a safety component of a product (or is itself a product) covered by the EU laws listed in Annex I, and that product needs third-party conformity assessment. Section A of Annex I still includes, for example, lifts, equipment for explosive atmospheres, pressure equipment, radio equipment, toys and medical devices. These apply from 2 August 2028.

Route 2, Annex III. Listed use cases, from 2 December 2027. Two matter in industry: AI used as a safety component in the management and operation of critical infrastructure (water, gas, heating and electricity supply, road traffic and critical digital infrastructure), and AI used in employment and worker management, such as allocating tasks or monitoring and evaluating staff performance. Under Article 6(3), an Annex III system is not high-risk if it does not pose a significant risk, for example because it performs only a narrow procedural task, but you must document that reasoning and register.

Here is how common industrial uses tend to land. Treat this as a first sort, then check the details with your compliance team:

What does moving machinery to Section B mean?

Under the original AI Act, an AI safety component in a machine could face two regimes at once: the AI Act's high-risk rules and the Machinery Regulation (EU) 2023/1230. The omnibus resolves this by moving the Machinery Regulation from Section A to Section B of Annex I, so most AI Act high-risk provisions no longer apply directly to machinery.

Instead, the Commission is empowered to adopt delegated acts adding the AI Act's high-risk requirements (Chapter III, Section 2 and Articles 17, 19, 72 and 73) to the essential health and safety requirements in Annex III of the Machinery Regulation, applying by 2 August 2028. Until machinery-specific standards exist, AI Act harmonised standards can give a presumption of conformity under Article 20 of the Machinery Regulation. Machine builders then do one conformity assessment, under the Machinery Regulation.

The more immediate date for machine builders is 20 January 2027, when the Machinery Regulation replaces the Machinery Directive. Its Annex I Part A already lists safety components with fully or partially self-evolving behaviour using machine learning that ensure safety functions, and machinery embedding such systems, as categories that need third-party (notified body) conformity assessment. If your machine uses a learned model to perform a safety function, that is the rule to plan for now.

What should engineers using or building AI in industry do now?

1. Make an inventory. List every AI system you build, buy or operate: vision cells, anomaly detectors, maintenance chatbots, scheduling tools, AI features inside vendor software.
2. Classify each one with the steps above and keep the reasoning on file. Most will be minimal-risk or Article 50 only, and writing that down now saves work later.
3. Fix Article 50 today. Any assistant or chatbot that talks to staff or customers needs a clear AI disclosure.
4. Keep safety functions conventional unless you are ready for a notified body. A light curtain, safety scanner and safety PLC remain the straightforward way to meet safety requirements. If AI improves productivity next to them, keep it out of the safety function. See safety PLCs, SIL and SIS for the conventional route.
5. Build traceability in. Version your training data, models and test results, and log what the model did in production. The high-risk rules ask for this, and so does any sensible quality system.
6. Design human oversight. Operators should understand what the AI does, see when it is uncertain and be able to override it or switch it off.

How to build the skills

The free courses on EDWartens cover both the technology and the governance:

• ISO 42001 and AI Governance for AI management systems, risk assessment and the AI Act's structure.
• Machine Vision and Quality Inspection for building and validating inspection systems.
• Predictive Maintenance with Machine Learning for the most common industrial AI use.
• AI for Industrial Automation for where AI fits around PLCs and SCADA.

For background, read AI quality inspection in manufacturing and predictive maintenance with AI, and browse all free AI courses for engineers. If you build LLM assistants, the AI Security and OWASP Top 10 for LLMs course covers the security side. The courses are free in full; the optional EDWartens Certificate of Completion starts from US$2.99 and is not a legal or ISO certification.

Take the free course

AI and machine learning · Intermediate · Free

ISO 42001 and AI Governance

AI and machine learning · Intermediate · Free

Machine Vision and Quality Inspection

AI and machine learning · Intermediate · Free

Predictive Maintenance with Machine Learning

AI and machine learning · Intermediate · Free

AI for Industrial Automation