Key Takeaways
- The Digital Omnibus on AI, Regulation (EU) 2026/1744, entered into force on 27 July 2026 and reset the timeline of the EU AI Act for humanoid robots deployed in the EU (Hunton Andrews Kurth, 2026).
- Three dates now matter most: transparency obligations apply since 2 August 2026, the Machinery Regulation applies from 20 January 2027, and high-risk AI embedded in machinery gets until 2 August 2028.
- The old 2 August 2026 high-risk deadline is gone. Stand-alone high-risk systems under Annex III now have until 2 December 2027, embedded AI in regulated products until 2 August 2028 (Orrick, 2026).
- The Omnibus points toward one conformity route for machines: the Commission can limit AI Act application where the Machinery Regulation covers the same ground, but the delegated act that settles this is only due by 2 August 2028. Until then, plan for both.
- Under the Machinery Regulation, AI safety components cannot be self-assessed for CE marking. A notified body must be involved (Nemko Digital, 2026).
If your compliance plan still says the EU’s high-risk AI rules hit humanoid robots on 2 August 2026, it has been out of date since 27 July. On that day the Digital Omnibus on AI entered into force and rewrote the EU AI Act for humanoid robots, moving the deadlines that matter for anyone deploying humanoid robots in manufacturing, warehousing or logistics inside the EU. Most robotics coverage I checked still cites the old dates.
I read the law-firm analyses of the final Omnibus text and the Machinery Regulation guidance rather than the headlines, and this post lays out the timeline as it now stands: what already applies, what moved, and what to demand from vendors before January 2027. One sentence of housekeeping first: I am a publisher, not a lawyer, so treat this as sourced analysis and not legal advice.
Table of Contents
What Is the EU Digital Omnibus and Why Does It Matter for Robotics?
The Digital Omnibus on AI is a simplification package that amends the AI Act. It was published as Regulation (EU) 2026/1744 and entered into force on 27 July 2026, three days after publication in the Official Journal (Hunton Andrews Kurth, 2026). For robot buyers and builders it does two big things: it postpones the high-risk compliance deadlines, and it starts merging the AI conformity route for machines with the route that machinery already follows for CE marking.
That second part is the one robotics coverage keeps missing. A humanoid robot is, in EU legal terms, machinery with embedded AI. Before the Omnibus, it was heading for two parallel compliance tracks: one under the AI Act, one under the Machinery Regulation. The Omnibus moved the Machinery Regulation from Section A to Section B of the AI Act’s Annex I and gave the Commission the power to limit AI Act application where sectoral legislation contains equivalent AI-specific requirements (Orrick, 2026). That is a routing and deduplication mechanism, not an exemption, and I come back to what it does and does not mean below.
The Omnibus is not purely a loosening. It also added new prohibited practices with their own near-term deadline, and it left the penalty ceiling untouched: up to EUR 35 million or 7 percent of global annual turnover for prohibited-practice violations under Article 99 of the AI Act.
What Does the EU AI Act for Humanoid Robots Require Today?
The transparency obligations in Article 50 apply since 2 August 2026 (Hunton Andrews Kurth, 2026; White & Case, 2026). Three duties are live now: AI systems interacting with people must disclose that the user is dealing with AI, synthetic content must be marked as AI-generated, and deepfakes must be labelled.
For humanoid robots the first duty is the practical one. If your robot carries a conversational interface, and nearly every current humanoid platform does, the people it talks to must be able to recognise they are speaking with an AI system. In a factory that is mostly a signage and onboarding question. In customer-facing or home settings it is a product design question, and it is worth putting to your vendor in writing today, because the obligation is already in force. Separate obligations for general-purpose AI models sit with the model providers, not with the factory deploying a robot, so they matter for your vendor’s supply chain rather than for your site.
The next date is 2 December 2026. From that day the AI-generated content marking obligations extend to systems that were already on the market before 2 August 2026, and the new prohibitions added by the Omnibus take effect: AI systems that generate child sexual abuse material or nonconsensual intimate imagery are banned outright (Hunton Andrews Kurth, 2026; Orrick, 2026). The prohibitions will rarely touch an industrial deployment, but the marking extension closes the grandfathering gap for older systems, so “we bought it before August” stops being an answer.
When Do the High-Risk AI Rules Actually Take Effect?
Two dates now define the high-risk layer of the EU AI Act for humanoid robots, depending on what kind of system you have. Stand-alone high-risk AI systems under Annex III must comply by 2 December 2027, moved back from the original 2 August 2026. High-risk AI embedded in products that are already regulated under Annex I, which covers machinery, medical devices and toys, gets until 2 August 2028 (Hunton Andrews Kurth, 2026; Orrick, 2026).
For a humanoid robot the embedded track is usually the relevant one. The AI that plans motion, avoids collisions and keeps the machine within its safety envelope is AI embedded in machinery, so its high-risk obligations land on 2 August 2028. The 2 December 2027 date matters where a deployment also uses stand-alone Annex III systems around the robot, for example AI used in employment contexts such as monitoring or evaluating workers. If your robot program includes workforce analytics, that part of the stack runs on the earlier clock.
Here is the full sequence as it now stands, with the Machinery Regulation date included, because for robot deployments the two timelines only make sense together.
Note what this chart quietly corrects: any article, vendor slide or internal memo that still names 2 August 2026 as the high-risk deadline is describing the pre-Omnibus law. I found that wrong date in most robotics-focused coverage I reviewed this week. The postponement is real, but so is the trap in it: the nearest hard date for robot deployments is no longer an AI Act date at all. It is the Machinery Regulation on 20 January 2027.
How Does the Machinery Regulation Fit Into the AI Act Timeline?
The Machinery Regulation (EU) 2023/1230 applies from 20 January 2027. It replaces the Machinery Directive 2006/42/EC and, as a regulation, applies directly in every member state without national transposition (EU-OSHA, 2026). For humanoid robots it is the nearer and, in practice, the harder deadline, and I keep its full requirements out of this post deliberately: my EU Machinery Regulation checklist walks through the compliance work item by item. This post covers the AI Act side and how the two now interlock.
The interlock is the Omnibus sector-overlap mechanism. The Machinery Regulation moved from Annex I Section A to Section B of the AI Act, and the Commission may limit AI Act application where sectoral legislation contains equivalent AI-specific requirements. A delegated act has to clarify these rules by 2 August 2028 (Orrick, 2026). The precise reading matters: the AI Act still applies to machinery. What the mechanism enables is routing, so that a robot maker does not prove the same thing twice under two laws. Until the delegated act lands, nobody can tell you exactly where the line will sit.
The Omnibus also narrowed the definition of a safety component to parts whose intended purpose is preventing or mitigating risks to health and safety, which trims how much of a robot’s software stack falls under the strictest rules (Orrick, 2026).
What is already certain is what the Machinery Regulation itself demands for AI. AI-based safety components cannot be self-assessed for CE marking: notified body involvement is mandatory for the Annex I Part A categories (Nemko Digital, 2026). The regulation also requires cybersecurity protection of safety functions, a safety review of AI and self-evolving behaviour, permits digital documentation and a digital Declaration of Conformity, and sets a 10-year documentation retention period (Nemko Digital, 2026; ABB, 2026).
| Question | AI Act (post-Omnibus) | Machinery Regulation |
|---|---|---|
| When does it bite for a humanoid robot? | Transparency now; embedded high-risk obligations by 2 August 2028 | 20 January 2027, directly in all member states |
| Who assesses AI safety conformity? | To be settled by delegated act, due by 2 August 2028 | Notified body; AI safety components cannot be self-assessed |
| What does it add beyond classic machine safety? | Risk management, data governance, human oversight for high-risk AI | Cybersecurity of safety functions, review of self-evolving behaviour, 10-year documentation |
| Biggest open question | How far the Commission limits AI Act application for machinery | Notified body capacity for AI safety components before 2027 |
Stay ahead of the humanoid robot market
Three briefings a week on deployments, pricing, and safety rules, written for operations leaders who need decisions, not hype. Free, no spam, unsubscribe anytime.
What Should Manufacturing Leaders Do Before 2027?
Turn the vendor conversation into a written procurement gate. The dates above are the vendor’s problem to solve and your problem to verify, and January 2027 is close enough that the answers should exist today. Five questions belong in every RFQ and every renewal:
- Which conformity route are you following? Machinery Regulation from 20 January 2027, and how the vendor plans to handle the AI Act’s embedded high-risk obligations by 2 August 2028. “We are waiting for the delegated act” is an acceptable answer for the routing detail, but not for the Machinery Regulation date.
- Which parts of the robot are AI safety components? Under the narrowed definition, that means parts whose intended purpose is preventing or mitigating risks to health and safety. Each one needs a notified body in the conformity assessment. Ask which notified body, and whether the engagement has started.
- Show me the CE Declaration of Conformity plan. For machines shipping after 20 January 2027, a DoC referencing the old Machinery Directive is the wrong document.
- How are safety functions protected against cyber attack, and how is self-evolving behaviour reviewed? Both are explicit Machinery Regulation duties, not optional extras (ABB, 2026).
- What documentation do I get, and for how long? The 10-year retention duty is the vendor’s, but your own records need to match. My robot safety documentation guide covers what to keep on your side of the file.
None of this replaces the technical safety work. The standards layer, from ISO 10218:2025 to the emerging humanoid robot safety standards, is still how you demonstrate the machine is actually safe. The regulations decide who must prove it, to whom, and by when.
What Does This Mean for Robots Already Deployed or Ordered?
Nothing in the Omnibus retroactively bans a deployed robot. The obligations attach at defined dates, and the postponements bought everyone time. But three things reach machines that are already on your floor or in your order book.
The marking extension arrives first. From 2 December 2026, AI-generated content marking obligations cover systems that were already on the market before 2 August 2026 (Hunton Andrews Kurth, 2026). If an installed system generates synthetic content, its age does not exempt it.
The Machinery Regulation applies to machines placed on the market from 20 January 2027. A robot delivered in 2026 under the Machinery Directive stays legal. The same model delivered in February 2027 needs Machinery Regulation conformity. For multi-year rollouts this splits one fleet across two legal regimes, which is a documentation and audit headache worth planning for now.
The fine print will keep moving. The delegated act on the machinery overlap is due by 2 August 2028, and until it arrives the exact division of labour between the two regulations stays open (Orrick, 2026). My working assumption: keep both tracks in the contract language and let the delegated act delete work later, rather than assuming deduplication that has not been enacted yet. The deadlines moved once. Building a compliance plan that survives them moving again is the cheapest insurance available.
Frequently Asked Questions
Does the EU AI Act apply to humanoid robots?
Yes. The transparency duties of the EU AI Act for humanoid robots apply now under Article 50, and high-risk obligations for AI embedded in machinery apply from 2 August 2028. The open question is not whether the AI Act applies but how much of it will be routed through the Machinery Regulation’s conformity process, which a delegated act must clarify by 2 August 2028 (Orrick, 2026).
Did the Digital Omnibus weaken the AI Act?
It postponed the high-risk deadlines and narrowed the safety component definition, but it also added new prohibited practices effective 2 December 2026 and extended content marking to older systems. The penalty ceiling of EUR 35 million or 7 percent of global turnover for prohibited practices under Article 99 of the AI Act is unchanged.
Do I need separate AI Act and CE conformity assessments for a robot?
The direction of travel is one route: the Omnibus lets the Commission limit AI Act application where the Machinery Regulation imposes equivalent AI-specific requirements. Until the delegated act settles the details, plan for Machinery Regulation conformity with notified body involvement for AI safety components, and treat any remaining AI Act obligations as additions to that file, not a parallel universe.
What happens if a vendor cannot produce a Declaration of Conformity?
A machine without a valid DoC cannot be placed on the EU market, and from 20 January 2027 that DoC must reference the Machinery Regulation for newly placed machines (EU-OSHA, 2026). Treat a missing or stale DoC plan as a disqualifying answer in procurement, not a detail to fix after signature.
Sources
- Hunton Andrews Kurth, “EU Digital Omnibus on AI Enters Into Force”, Privacy and Cybersecurity Law blog, 2026. Article
- Orrick, “EU AI Act Update: Digital Omnibus Finalizes 8 Compliance Changes”, July 2026. Insight
- White & Case, “EU Agrees Digital Omnibus Deal to Simplify AI Rules”, 2026. Insight alert
- EU-OSHA, “Regulation (EU) 2023/1230 on machinery”, legislation overview. Overview
- Nemko Digital, “EU Machinery Regulation: AI and Conformity Assessment”, 2026. Guide
- ABB, “EU Machinery Regulation 2027: What Changes for Machine Builders”, 2026. Overview
Last verified August 15, 2026. All dates were checked against the named legal analyses on the day of writing. This article is sourced analysis, not legal advice.
Ulrich Baldauf is the founder of There’s A Robot For That, covering humanoid robotics for manufacturing and industrial operations. He has tracked the humanoid robot sector since 2024, with a focus on safety standards (ISO 10218, EU Machinery Regulation 2023/1230) and what deployments mean for operations and EHS teams. Connect on LinkedIn: linkedin.com/in/ubaldauf
