Digital Omnibus AI: The Council's Letter to the European Parliament Explained

When the Council of the European Union formally transmits a negotiating position to the European Parliament, it does so through a procedure grounded in Article 294 of the Treaty on the Functioning of the European Union (TFEU). This is not merely administrative correspondence — it marks the official entry into the co-decision process known as the ordinary legislative procedure, and for the Digital Omnibus package on artificial intelligence, that letter carries enormous practical weight for every business timing its EU AI Act compliance investments.

What the Council's Transmittal Letter Means Under Article 294 TFEU

Article 294 TFEU lays out the ordinary legislative procedure in three potential readings. When the Council adopts a position at first reading and transmits it to the Parliament, the clock begins on the Parliament's three-month response window. During that period, Parliament may approve the position (making the act adopted as amended), reject it (requiring a qualified majority and blocking the text), or propose amendments of its own.

For the Digital Omnibus AI regulation — which amends Regulation 2024/1689, the EU AI Act — the Council's letter to Parliament represents the output of months of technical working group negotiations. The Council's general approach diverges from the Commission's original proposal in ways that businesses across all sectors must understand, because the trilogue negotiations that follow will produce the final binding text.

The formal transmittal creates a specific institutional dynamic: the rapporteur in the European Parliament's IMCO and LIBE committees, along with the shadow rapporteurs from each political group, now carry the mandate to negotiate with Council representatives and Commission officials in closed trilogue sessions. The letter therefore effectively freezes the Council's starting position while opening the floor to Parliament's counter-proposals.

Article 6: High-Risk Classification and What the Council Changed

The Commission's original text in Article 6 established a two-tier classification system: AI systems listed in Annex I that are safety components of regulated products, and AI systems performing the tasks listed in Annex III across eight domains including biometric identification, critical infrastructure, education, employment, essential services, law enforcement, migration, and administration of justice.

The Council's position introduces a materiality threshold into the Article 6(2) classification. Rather than any AI system performing an Annex III task being automatically high-risk, the Council text requires an assessment of whether the system poses a significant risk of harm to health, safety, or fundamental rights. This means that a narrowly scoped AI tool used, for example, in creditworthiness assessments under Annex III point 5(b) may not automatically qualify as high-risk if the deployer can demonstrate the system's output is merely one factor among many in the final human-led decision.

For businesses, this shift matters because it creates a classification grey zone. Where the Commission's text pushed toward a presumption of high-risk status, the Council's position allows for a documented self-assessment process that could keep many enterprise AI deployments outside the most burdensome compliance track. However, this also increases documentation obligations — the justification for a non-high-risk determination must itself be auditable.

Article 43: Conformity Assessment Routes Under the Council Position

Article 43 governs how providers of high-risk AI systems demonstrate compliance. The Commission proposed that most high-risk systems could self-certify through internal controls (Annex VI), with mandatory third-party assessment only for biometric identification systems and certain other sensitive categories.

The Council's position preserves this general architecture but adds nuance around Article 43(1)(b), which concerns AI systems already subject to third-party assessment under harmonised legislation (such as medical devices under the MDR or machinery under the Machinery Regulation). The Council text clarifies the relationship between existing notified body assessments and the incremental AI Act requirements, reducing the risk of duplicative conformity assessment for products already in regulated product categories.

This clarification is particularly valuable for manufacturers of medical devices incorporating AI diagnostic functions, where both the MDR conformity pathway and the AI Act requirements would otherwise potentially require separate notified body engagements.

Article 51: GPAI Model Obligations and the Council's Lighter Touch

General Purpose AI (GPAI) model obligations under Article 51 were one of the most contested provisions in the original Commission text. The Council's position adopts a tiered approach based on the training compute threshold (10^25 FLOPs for systemic risk designation) while simplifying the baseline obligations that apply to all GPAI models.

Compared to the Commission's draft, the Council removes several of the most operationally burdensome requirements for lower-tier GPAI models, including certain granular training data documentation requirements, while maintaining the core transparency obligations and the technical documentation package that providers must make available to downstream deployers. For systemic risk GPAI models, the Council maintains the adversarial testing and incident reporting obligations but adjusts the timing of Commission-issued standardised evaluation protocols.

Compliance Investment Timing Implications

The trilogue process typically takes six to eighteen months from the Council's first reading position. For businesses planning AI Act compliance investments, the key insight from the Council's letter is that the final text will almost certainly be closer to the Council's position than to the Commission's original proposal — the Parliament will secure amendments in certain priority areas (likely fundamental rights safeguards and enforcement coordination) but is unlikely to reverse the Council's structural concessions on Article 6 classification thresholds and Article 43 conformity assessment.

Businesses that began compliance programmes based on the original Commission text should model their timelines around the Council's position rather than waiting for the final text. The provisions that are most likely to survive trilogue intact — the risk management system requirements of Article 9, the logging obligations of Article 12, and the human oversight requirements of Article 14 — are well-defined enough to begin implementation work now. Investments in those areas will not be wasted regardless of how the remaining contested provisions resolve.

Frequently Asked Questions

What is the Digital Omnibus AI regulation and how does it differ from the original EU AI Act?

The Digital Omnibus AI package is a legislative proposal that amends the original EU AI Act (Regulation 2024/1689). It introduces targeted changes to provisions including the AI system definition, high-risk classification thresholds, conformity assessment procedures, and GPAI model obligations, generally moving toward a lighter regulatory burden for lower-risk deployments while preserving core safeguards.

How long will the trilogue process take before a final text is adopted?

Trilogue negotiations typically last between six and eighteen months, depending on the political complexity of the outstanding issues. Given the technical nature of many AI Act amendments, businesses should plan for a process lasting at least twelve months from the Council's formal transmittal to Parliament before a final text is available.

Should businesses pause compliance investments until the Digital Omnibus AI final text is published?

No. Core obligations under the original AI Act — including risk management systems (Article 9), technical documentation (Article 11), logging (Article 12), and human oversight (Article 14) — are not substantively contested in the Digital Omnibus negotiations. Investments in these foundations are safe to make now and will remain valid under any likely final text.

Sources

• Treaty on the Functioning of the European Union, Article 294 (Ordinary Legislative Procedure)
• Regulation (EU) 2024/1689 of the European Parliament and of the Council (EU AI Act), Articles 6, 9, 11, 12, 14, 43, 51
• Council of the European Union, General Approach on the Digital Omnibus AI Regulation, Council Document 2025
• European Commission, Proposal for a Regulation amending Regulation (EU) 2024/1689 (Digital Omnibus), COM(2025)
• European Parliament, IMCO/LIBE Committee Working Documents on AI Act Amendment