The AI Omnibus Trilogue Collapsed. Your August Deadline Is Back — For Now.

What Actually Happened on April 28

The final political trilogue on the EU AI Omnibus concluded after 12 hours of negotiations without a deal.^[1] EU member states and European Parliament lawmakers remain apart on fundamental questions, and talks will resume next month.

This is a significant development. Just days earlier, legal analysts were projecting agreement by end of April, with formal adoption by July — ahead of the August 2 deadline. The breakdown resets those expectations.

At the core of the disagreement: whether companies operating in sectors already covered by existing EU product safety rules should be exempted from the AI Act's high-risk requirements altogether.^[1] Some member states and Parliament members want carve-outs for industries already subject to sectoral regulation — including AI used in biometrics, utilities supply, health, creditworthiness, and law enforcement contexts. The Parliament has resisted those exemptions, arguing that existing sectoral rules do not adequately cover AI-specific risks.^[2]

What This Means for Your August 2 Deadline

The delay to the AI Omnibus has not yet been published in the Official Journal. Until it is, the original AI Act text remains in force. That means August 2, 2026 is still your legal deadline for Annex III high-risk AI obligations.^[1]

To be clear: the trilogue failure does not mean the delay is cancelled. It means the delay has not yet been formally adopted. If the institutions reach agreement in May and the text is adopted and published before August 2, the delay takes effect before your deadline arrives. But if the May negotiations also fail or are delayed, the original August 2 date stands.

The safest position right now: treat August 2, 2026 as your operative deadline. Do not plan around a delay that does not yet exist in law.

The Broader AI Act Is Still Coming — And Universities Are Already Feeling It

While trilogue negotiators argue about exemptions and scope, the AI Act's effects are already landing on organisations that did not plan for them. A new report from Times Higher Education published April 27 quotes the European University Association's director of policy coordination warning that European universities may have to "change everything" about how they use AI once the high-risk obligations take effect.^[3]

The concern is specifically about AI used for student assessment. Under Annex III of the AI Act, AI systems used to evaluate learning outcomes and assess the appropriate level of education an individual will receive are classified as high-risk.^[4] That classification covers AI tools used to grade assignments, infer student performance levels, or direct learners into particular educational pathways.

Here is the part that is already creating compliance anxiety across European campuses: academics are already using tools like ChatGPT to assess student work — in many cases without formal institutional oversight. If those tools fall under the AI Act's definition of high-risk systems, the institutions deploying them are subject to the full high-risk compliance regime: conformity assessments, technical documentation, human oversight requirements, record-keeping, and AI literacy obligations for staff.^[3]

The regulatory guidance from the EU AI Office on how these obligations apply to informal AI use in educational contexts is still awaited, which means institutions are currently operating in a compliance grey zone.^[3] That uncertainty does not pause the deadline.

The AI Literacy Question — Binding or Optional?

One of the three unresolved issues in the trilogue that CDT's bulletin flags is AI literacy obligations.^[2] The Council wants to replace the AI Act's existing binding AI literacy requirement with a non-binding encouragement. The Parliament has kept the binding duty on providers and deployers to maintain AI literacy programmes for their staff.

For SMEs: if your AI literacy training programme is currently being deprioritised on the assumption that AI literacy obligations are likely to become optional — that assumption depends on the Council's position winning in the trilogue. Right now, that outcome is not certain. Until the final text is agreed and published, your AI literacy programme remains a binding obligation under the current AI Act text.

What to Watch For in May

The trilogue resumes next month. Three things to track:

Did the institutions agree on a new date? If May trilogue produces a deal, the timeline for Official Journal publication remains achievable before August — but only if the agreement is reached quickly.
What happened to the sectoral exemption fight? If the Parliament gives ground on high-risk AI embedded in regulated products, fewer AI systems qualify as high-risk. That reduces compliance scope for a range of AI-enabled products. If the Council gives ground, the exemptions disappear and the original scope holds.
Is the AI literacy obligation binding or optional? This determines whether you need documented AI literacy programmes for your staff or whether you can treat training as a soft recommendation.

The Bottom Line for SMEs

The Omnibus delay is not law yet. Your August 2, 2026 deadline stands until the Official Journal says otherwise. The trilogue collapse does not give you more time — it gives you more uncertainty, which is a different problem.

For SMEs in the education sector, or companies selling AI assessment tools into European universities: the high-risk classification of student assessment AI is not in dispute. Your compliance obligations under Annex III are approaching regardless of what happens in the trilogue. If you have not started your conformity assessment process, the clock is still running.

We will update this analysis as soon as the May trilogue produces an outcome.

This article is for informational purposes only and does not constitute legal advice.