Digital Omnibus: What Actually Changes in the EU AI Act

The Digital Omnibus to the AI Act has applied since 27 July 2026. The headlines spoke of relief and delay. That is true for some of the rules and not for the others. The Digital Omnibus postpones the deadlines for high-risk systems and words the AI literacy duty more softly. The transparency duties and the oversight of models that have applied since 2 August 2026 stay in place. For companies that only use AI and do not build it, the relief is therefore smaller than it sounds.

27 Jul 2026

The Digital Omnibus is in force, published on 24 Jul 2026

Regulation (EU) 2026/1744

2 Dec 2026

End of the transition period for labelling systems already on the market

AI Act, Art. 113

2 Dec 2027

New date for high-risk systems under Annex III, Annex I follows on 2 Aug 2028

AI Act, Art. 113

Contents

• What the Omnibus is and is not
• The new deadlines at a glance
• What changes
• What stays
• What deployers do now
• Conclusion: the delay applies to high-risk, not to the rest
• Frequently asked questions about the Digital Omnibus
• Sources

What the Omnibus is and is not

The Digital Omnibus is Regulation (EU) 2026/1744 of 8 July 2026. It was published in the Official Journal on 24 July 2026 and has been in force since 27 July 2026 (Sources: EUR-Lex; European Commission, 2026). It amends the AI Act, the EASA Regulation and the Machinery Regulation. It is not a new rulebook: the AI Act remains, with its risk classes, prohibitions and transparency duties.

What the Omnibus does not do matters just as much. It does not abolish the duties for high-risk systems, it postpones their start. It does not delete AI literacy, it rewords it. And it leaves the transparency duties of Article 50 untouched, which have applied since 2 August 2026 (Source: AI Act Service Desk, timeline).

The new deadlines at a glance

Four dates matter for companies. Two of them already apply or are imminent, two have been pushed back.

For presentations:SVGPNG

The rules for high-risk systems under Annex III, for example in HR, education or credit scoring, now apply from 2 December 2027. Systems embedded in regulated products under Annex I follow on 2 August 2028 (Source: AI Act, Art. 113 as amended by Regulation (EU) 2026/1744). For labelling, a short transition period applies: providers of systems that generate synthetic audio, image, video or text content and were placed on the market before 2 August 2026 must comply with Article 50(2) by 2 December 2026 (Source: AI Act, Art. 113).

What changes

High-risk is postponed. Anyone running or planning a high-risk system gains a good year. That is time for an inventory, documentation and vendor checks, not an all-clear. Deployers who are only planning such a system can handle selection and contracts more carefully instead of deciding under deadline pressure.

AI literacy is worded more softly. Under the new Article 4, providers and deployers must “take measures to support the development of AI literacy”. The duty no longer requires them to guarantee a specific level (Source: Regulation (EU) 2026/1744, Art. 4 as amended). Responsibility for the framework and for promotion also shifts further towards the Commission and Member States (Source: European Commission, 2026).

SME relief now covers small mid-caps. Simplified duties, for example for technical documentation, no longer apply only to small and medium-sized enterprises but also to companies just above those thresholds (Source: Regulation (EU) 2026/1744). For German mid-sized companies, that is practically relevant.

New prohibitions are added. AI systems that generate non-consensual intimate depictions of real people or child sexual abuse material are prohibited from 2 December 2026 (Sources: Regulation (EU) 2026/1744, Art. 113; AI Act Service Desk). For most companies this changes nothing, but anyone who makes image generators available should know it.

Oversight is bundled. The AI Office receives exclusive competence for AI systems built on general-purpose models from the same provider (Source: Regulation (EU) 2026/1744). Registration in the EU database is streamlined, and processing special categories of personal data to detect and correct bias is permitted (Source: European Commission, 2026).

What stays

Since 2 August 2026, enforcement of the AI Act has started, at national and EU level. It covers general-purpose models, prohibitions, transparency rules and AI literacy (Source: AI Act Service Desk, timeline). The duties for providers of general-purpose models have applied since 2 August 2025.

For deployers, that means:

• Labelling: Article 50 applies. Anyone who publishes AI-generated content or lets AI systems interact with people is affected. What exactly it requires is covered in our article on AI content disclosure rules.
• Prohibitions: The prohibited practices have applied since February 2025 and have not been postponed.
• AI literacy: Worded more softly, but still a duty. Anyone who takes no measures has nothing to show for it.

What deployers do now

For companies that use AI systems, three steps make sense, regardless of the risk class their use falls into. They cost little, and each one feeds several duties at once.

1. Build an inventory. Which AI systems run in the company, who uses them, on which data and for what purpose? Without this list, no risk class can be determined. It includes tools that individual teams introduced themselves.

2. Assign roles. Who is accountable for the use of AI, who decides on new systems, who vets vendors? This is the core of governance for AI agents, and it matters more with every additional system.

3. Check literacy and labelling. Are there training sessions and guidelines you can document? And do you label AI-generated content the way Article 50 requires? For systems that generate synthetic content and were introduced before 2 August 2026, the transition period ends on 2 December 2026.

You will find the terms explained in the glossary under EU AI Act.

Conclusion: the delay applies to high-risk, not to the rest

The Digital Omnibus gives companies with high-risk systems more time and mid-sized companies some relief. For most deployers, little changes day to day: labelling, prohibitions and AI literacy apply, and oversight is running.

Anyone who waits now because “everything was postponed” confuses two things. What was postponed is the start for high-risk. The work this requires, inventory, roles, documentation, does not get smaller, it just gets more lead time.

How to set this up for your use of AI is something we map out in a free diagnosis call.

Frequently asked questions about the Digital Omnibus

What is the Digital Omnibus on the AI Act?

It is Regulation (EU) 2026/1744, which amends the AI Act. It has been in force since 27 July 2026, postpones the deadlines for high-risk systems and simplifies individual duties (Sources: EUR-Lex; European Commission, 2026).

When do the rules for high-risk systems apply now?

Systems under Annex III apply from 2 December 2027, systems under Annex I from 2 August 2028 (Source: AI Act, Art. 113).

Do I still have to ensure AI literacy?

Yes, the duty remains but is worded more softly. Companies must take measures to support AI literacy, and they do not have to guarantee a specific level (Source: Regulation (EU) 2026/1744, Art. 4 as amended).

Does the labelling duty apply yet?

Yes, Article 50 has applied since 2 August 2026. For systems that generate synthetic content and were placed on the market before that date, the transition period for Article 50(2) ends on 2 December 2026 (Source: AI Act, Art. 113).

What changes for mid-sized companies?

The relief for small and medium-sized enterprises also covers small mid-caps, for example for technical documentation (Source: Regulation (EU) 2026/1744).

Sources

• European Union, 2026: Regulation (EU) 2026/1744 (Digital Omnibus on AI), Official Journal of 24 Jul 2026
• European Commission, 2026: AI Omnibus enters into force
• European Commission, AI Act Service Desk: Timeline for the implementation of the EU AI Act
• European Commission, AI Act Service Desk: Article 113 AI Act, entry into force and application