Law & Compliance · 2026-10-04 (Last updated: 9 November 2026) · 12 min read
EU Digital Omnibus: what the simplification package changes in the AI Act, who gains relief as a small mid-cap and why waiting is still risky

Michael Kaiser
Co-Founder & Head of Systems, Vincency
The short answer first: the AI part of the Digital Omnibus is no longer a proposal. It has been law since 27 July 2026, as Regulation (EU) 2026/1744 of 8 July 2026. For mid-market companies it changes two things above all. The AI Act’s high-risk duties start later, but on fixed dates: from 2 December 2027 for systems under Annex III and from 2 August 2028 for AI in products under Annex I. And the simplified documentation that used to be reserved for SMEs now also covers small mid-caps: companies with fewer than 750 employees and an annual turnover not exceeding 150 million euros or a balance sheet total not exceeding 129 million euros. Everything that already applied still applies: the bans on unacceptable AI practices, the duties for general-purpose models and the transparency duties under Article 50.
This piece is written for the managing director who read about Brussels easing AI rules and now wonders whether the compliance project can be shelved. The honest answer is no, and the details explain why.
What Regulation (EU) 2026/1744 actually changes
The Commission tabled the omnibus on 19 November 2025 as COM(2025) 836 final. The reasoning was practical: harmonised standards, guidance and national supervisory structures were arriving slower than the AI Act’s application dates, so companies would have faced obligations without the instruments to meet them. The proposal therefore wanted to tie the start of the high-risk rules to the availability of those instruments. Parliament and Council did not adopt that mechanism and set fixed dates instead. For a company that is the better news: a date can be planned, a Commission confirmation at an unknown point in time cannot.
| Element | Until July 2026 | Since 27 July 2026 |
|---|---|---|
| High-risk under Annex III | 2 Aug 2026 | 2 Dec 2027 |
| High-risk in products under Annex I | 2 Aug 2027 | 2 Aug 2028 |
| Transparency duties, Art. 50 | From 2 Aug 2026 | Not postponed |
| Marking under Art. 50(2) for systems already on the market | 2 Aug 2026 | 2 Dec 2026 |
| Simplified technical documentation | SMEs only | SMEs and small mid-caps |
| AI literacy, Art. 4 | Ensure a sufficient level (obligation of result) | Take measures to promote it (obligation of means) |
| Registration of systems exempted under Art. 6(3) | Required | Remains, in simplified form; the proposed deletion was rejected |
| New bans | None | Sexualised deepfakes without consent and child sexual abuse material, from 2 Dec 2026 |
| Existing bans and GPAI duties | Applicable | Unchanged |
The new size class that decides who wins
The more structural change sits in the size categories. Commission Recommendation (EU) 2025/1099 of 21 May 2025 created a new tier: small mid-cap enterprises are companies that have outgrown the SME definition, employ fewer than 750 people and have an annual turnover not exceeding 150 million euros or a balance sheet total not exceeding 129 million euros. The Commission counts roughly 38,000 of them in the EU. Before the Omnibus, the moment a company passed 250 employees it fell off a cliff into the full large-enterprise regime. The omnibus uses the new category to carry the simplified technical documentation of SMEs upward.
For the German Mittelstand this is the concrete question worth checking: companies between 250 and 749 employees sit exactly in this band, precisely where the classic Mittelstand grows out of the small-enterprise bracket. The relief concerns providers, though, meaning companies that develop a high-risk system themselves or place it on the market under their own name. A company that only uses a purchased system has deployer duties, and those are set out in our AI Act deployer obligations guide.
What changes for machine builders
One change hits a core sector of the Mittelstand directly and barely appears in the summaries. Regulation (EU) 2026/1744 takes AI-enabled machinery out of the direct application of the AI Act’s high-risk regime: the Machinery Regulation moves from Section A to Section B of Annex I of the AI Act. Such products therefore follow the sectoral route of the Machinery Regulation. The Commission has to supplement its Annex III by delegated act with health and safety requirements for high-risk AI systems that are safety components of machinery or machinery themselves. What the Machinery Regulation already demands of software from 20 January 2027 is covered in our piece on the Machinery Regulation and software updates.
The wider package beyond the AI Act
The AI omnibus was one part of a broader digital package the Commission presented the same day, summarised in its official press release. Its second omnibus, proposal COM(2025) 837 on data and cybersecurity rules, is a separate legislative procedure and had not been adopted as of 9 October 2026. For operations the most relevant piece in it is a single entry point for cybersecurity incident reporting: until it exists, the same incident can trigger duties under NIS2, the Cyber Resilience Act and GDPR at once, a burden we documented in our pieces on the NIS2 reporting deadlines and the CRA duty that is live since September 2026. It also touches GDPR in targeted ways and modernises cookie rules toward browser-level consent. The package further includes a Data Union strategy and a proposal for European Business Wallets; the Commission estimates the simplifications could save five billion euros in administrative costs by 2029. Unlike the AI part, the rule here is still: until Parliament and Council agree, nothing changes.
The honest risk calculation
Three qualifications keep this sober. First, deferral is not abolition: for Annex III systems 2 December 2027 is only a little over a year away, and risk management, technical documentation and data quality cannot be caught up in the last quarter. Second, relief is not exemption: small mid-caps keep the same obligations, they just meet them with less paperwork. Third, the parts already in force were not postponed, Article 4 was only weakened, and 2 December 2026 adds new bans and ends the marking transition (details in our article on the labelling duty for chatbots). A company that paused its AI inventory work on headline news has gained nothing. The practical stance is the same one we take in the AI Act deployer obligations guide: map which systems qualify, prepare the documentation path, and read the time gained as borrowed time rather than abolished duty.
Whoever takes one sentence from this piece: the Digital Omnibus has stretched the clock and lowered the paperwork burden for part of the Mittelstand, but it changes the calendar, not the destination.
Free tool
Does this apply to you? Check in six questions
The EU regulation check asks about sector, size, products, AI use and invoicing and sorts which of five obligations apply to your company: NIS2, the AI Act, the Cyber Resilience Act, the Data Act and e-invoicing. A first assessment, not legal advice. No sign-up.
Run the checkRelated service
IT strategy as a guided service
The analysis described here can be done in house. If you lack the time or the distance from your own systems, we take it on: inventory, a prioritised order and, if you wish, the implementation.
IT strategy consultingFrequently asked questions about the Digital Omnibus
What is the EU Digital Omnibus exactly?
A package of two legislative proposals the Commission presented on 19 November 2025. The AI part, originally COM(2025) 836 final, has been in force since 27 July 2026 as Regulation (EU) 2026/1744. It amends the AI Act, Regulation (EU) 2024/1689, the Machinery Regulation (EU) 2023/1230 and Regulation (EU) 2018/1139 on civil aviation. The second part, COM(2025) 837 on data and cybersecurity rules with GDPR changes, cookie rules and a single entry point for incident reporting, is a separate procedure and had not been adopted as of 9 October 2026.
Does the omnibus postpone the AI Act deadlines of August 2026?
For the high-risk duties, yes, and to fixed dates. For high-risk systems under Annex III, such as AI used in recruitment or in evaluating employee performance, the duties apply from 2 December 2027; for AI in products under Annex I from 2 August 2028. The Commission had proposed tying the start to the availability of harmonised standards; Parliament and Council agreed on fixed dates instead. The transparency duties under Article 50 were not postponed and have applied since 2 August 2026. Only the machine-readable marking for systems already on the market before then has a transition period until 2 December 2026.
What is a small mid-cap and does my company qualify?
Commission Recommendation (EU) 2025/1099 defines small mid-cap enterprises as companies that no longer count as SMEs under Recommendation 2003/361/EC, employ fewer than 750 people and have an annual turnover not exceeding 150 million euros or a balance sheet total not exceeding 129 million euros. There are around 38,000 such companies in the EU. Anyone between 250 and 749 employees who stays below these financial ceilings used to fall under the large-enterprise rules. Since the omnibus, reliefs the AI Act previously granted only to SMEs are open to them.
Which reliefs do small mid-caps get concretely?
At the core, the simplified technical documentation that used to be reserved for SMEs. It concerns providers, meaning companies that develop a high-risk system themselves or place it on the market under their own name. A company that only uses a purchased system has deployer duties under Article 26 and draws up no technical documentation; for them the new size class changes nothing at this point. The obligations themselves remain; what drops is the formal effort of demonstrating compliance.
Can I stop my AI Act project now?
No. First, a large part already applies: the bans on unacceptable AI practices, the duties for general-purpose AI models and, since 2 August 2026, the transparency duties under Article 50. On 2 December 2026 two new bans follow, on sexualised deepfakes of identifiable people without consent and on child sexual abuse material. Second, 2 December 2027 is only a little over a year away for Annex III systems, and risk management, documentation and data quality cannot be built in the last quarter. Third, Article 4 remains: the AI literacy duty was weakened, not deleted.
Sources, status and note: Legal basis is Regulation (EU) 2026/1744 of 8 July 2026, published in the Official Journal on 24 July 2026. We checked the adopted content on 9 October 2026 against the briefing of the European Parliamentary Research Service of June 2026 (agreement of 7 May, plenary vote of 16 June, fixed dates, documentation relief for mid-sized companies, machinery rules) and the Council press release of 29 June 2026 on final adoption. Also used: the original proposal COM(2025) 836 final for comparison, Commission Recommendation (EU) 2025/1099 for the small mid-cap definition with fewer than 750 employees and the financial ceilings (point 2 of the annex, German and English versions retrieved on 8 November 2026), and the Commission press release for the wider digital package and the savings estimate. The Commission press release IP/25/1277 of 21 May 2025 for the figure of nearly 38,000 small mid-caps in the EU (retrieved on 16 October 2026). This article is an orientation, not legal advice.
Related insights