What changed, and what did not

According to law-firm summaries of the legislative record, the European Parliament adopted the Digital Omnibus on AI in mid-June 2026, the Council approved it at the end of June, and it entered into force in late July 2026. The headline change is timing: obligations for stand-alone high-risk systems listed in Annex III now apply from 2 December 2027, and for AI embedded in regulated products under Annex I from 2 August 2028.

Not everything moved. The transparency duties in Article 50 and the AI Office's enforcement powers over general-purpose AI providers took effect on 2 August 2026 as originally scheduled. If you have not yet looked at what those transparency rules require, our earlier article on them covers the practical changes.

A later date is not a lighter requirement. The substance of the high-risk obligations is unchanged; you have more time to build documentation, logging, and oversight properly.

Where to spend the extra months

The work that takes longest is rarely the legal reading. It is the engineering and organizational groundwork: knowing which of your systems are in scope, producing technical documentation that matches how the system really behaves, building event logging, and designing human oversight that a reviewer can genuinely exercise.

Starting now also gives you something valuable during enterprise sales. Buyers in regulated sectors are already asking suppliers how they will meet the high-risk rules, and a credible, dated plan answers that question better than a promise to check later.

WorkstreamWhy start nowOutput
Inventory and classificationYou cannot plan for systems you have not identifiedA list of AI systems with a provisional risk tier
Technical documentationIt must reflect real behavior, which takes iterationLiving documents per system
Logging and traceabilityRetrofitting logs into a running system is costlyDefined events, retention, access
Human oversight designOversight has to be usable, not nominalReview points with authority to reject
Supplier due diligenceYour obligations depend on what your providers give youQuestions and evidence requests sent

What not to assume

Do not assume the supporting technical standards you would build against are finished, and do not wait for them. Design to the stated requirements, keep your approach flexible, and adapt as guidance matures. Equally, do not treat the omnibus as the last word: this is an active area, and further simplification proposals are still moving through the EU institutions.

A practical rhythm is a short quarterly review that checks the official sources for changes, re-confirms your system inventory, and tracks progress on each workstream.

Want a plain-English AI Act readiness review for the systems you build or buy? Talk to our team.

Key takeaways

  • The Digital Omnibus on AI is in force: Annex III high-risk obligations now apply from 2 December 2027 and Annex I from 2 August 2028.
  • Article 50 transparency duties and AI Office powers over general-purpose AI were not delayed and applied from 2 August 2026.
  • The requirements did not get lighter; use the time for inventory, documentation, logging, oversight design, and supplier diligence.
  • Design to the requirements without waiting for every standard, and review official sources quarterly.
This is general information, not legal advice. Dates and details come from public regulator and law-firm summaries as of October 2026 and can change; confirm against the official text before relying on them.