Physical AI in the Digital Omnibus – Update

Co authors: Ramona Ahmadi, Julius Haas

What is the current status?

On May 6, 2026, the chief negotiators of the Council of the European Union and the European Parliament agreed on a provisional draft of the Digital Omnibus on AI [1], a legislative package aimed at simplifying the digital legal framework („Digital Omibus on AI“).[2]

Significant changes are expected, particularly for physical/industrial AI. This includes machines equipped with (high-risk) AI, such as autonomous machines and robots.

What changes are expected?

• New deadlines for the application of the rules on high-risk AI systems
  • The rules for stand-alone high-risk AI systems will take effect on December 2, 2027
  • For high-risk AI systems embedded in products, they will apply as of August 2, 2028.[3]
• Registration Requirements and Data Use

  In the future, providers will be required to register their AI systems in the EU database for high-risk systems – even if they themselves believe that their system is not a high-risk system. In addition, the processing of special categories of personal data for the purpose of detecting and correcting bias will only be permitted when absolutely necessary.[4]

• Regulatory sandboxes and transparency solutions

   National authorities must establish AI regulatory sandboxes by August 2, 2027. However, the deadline for providers to implement transparency solutions for synthetically generated content has been shortened to three months; the deadline is December 2, 2026.[5]

• Supervision by AI Office

The responsibilities of the Office for Artifical Intelligence regarding AI systems based on general-purpose AI models have been clarified. In certain cases, however, national authorities remain responsible.[6]

Finally, so-called “nudifier” apps are also to be banned under Article 5 of the AI Act. They are AI systems that generate non-consensual sexual or intimate content, as well as depictions of child sexual abuse. Companies have until December 2, 2026 to adapt their systems.[7] From that date onward, these will be banned under Article 5 of the AI Act.[8]

Digital Omnibus and Machines

To avoid overlaps between the EU Machinery Regulation (EU) 2023/1230 and the AI Regulation (Regulation (EU) 2024/1689), a more nuanced, sector-specific approach to regulating AI in machinery shall be adopted[9]:

• The EU Machinery Regulation is being removed from the direct scope of application of the AI Regulation and moved from Annex I, Part A, to Part B. This is because, pursuant to Article 2 Para. 2 of the AI Regulation, legal acts listed in Annex I, Part B are largely exempt from the AI Act. A more extensive transfer of additional product categories (e.g., toys or medical devices) was discussed but ultimately rejected.
• Instead, the European Commission is empowered to establish additional requirements for high-risk AI in machinery by means of delegated acts. To this end, the Commission shall be required to adopt delegated acts amending Annex III of the Machinery Regulation. In doing so, particular consideration shall be given to the requirements set forth in Chapter III, Section 2, as well as Articles 17, 19, 72, and 73 of the AI Regulation.
• The relevant delegated acts are to enter into force by August 2, 2028.

As a result, machines equipped with high-risk AI will be subject to sector-specific regulations in the future: The previous dual regulatory framework will be streamlined so that machines with AI are subject primarily to sector-specific safety regulations –  rather than being subject to both the AI Regulation and product safety legislation cumulatively. At the same time, an equivalent level of protection for health and safety is to be maintained.

Transition Period and Standards

During the transition period, manufacturers may rely on:

• harmonized standards or
• common specifications

developed on the basis of the AI Regulation and covering the requirements for the presumption of conformity under Article 20 of the EU Machinery Regulation (EU 2023/1230) – at least until specific AI standards for the Machinery Regulation are available.

Risk assessment (important for practical application!): Until harmonized standards are published, risk assessments must be conducted - even without referring to them - based on current technical standards; to avoid liability risks, the relevant state of science and technology must also be taken into account.

Practical Guidance

The Digital Omnibus Act clearly shifts the focus of industrial AI toward sector-specific product safety regulations. For manufacturers, this means

• greater individual responsibility in the short term (risk assessment without standards),
• less double regulation in the medium term
• but also stricter liability requirements that go beyond the current state of science and technology.