After 12 grueling hours of negotiations on Tuesday, EU member states and European Parliament lawmakers walked away from the table without a deal on proposed changes to the bloc's landmark AI Act. The collapse of the so-called Digital Omnibus trilogue — the three-way negotiations between the European Commission, Council, and Parliament — leaves the world's most ambitious AI regulation in limbo, with an August compliance deadline now bearing down on thousands of companies that had been counting on more time.
"It was not possible to reach an agreement with the European Parliament," a Cypriot official confirmed to Reuters, with Cyprus currently holding the rotating EU Council presidency. The next round of talks is now penciled in for May 13, giving negotiators just weeks to bridge what proved to be an unbridgeable gap on Tuesday.
The Product-Safety Fault Line
The central dispute that torpedoed the talks is deceptively narrow but carries enormous consequences: should AI systems embedded in products already regulated under existing EU safety rules — medical devices, toys, machinery, automobiles — be exempt from the AI Act's high-risk obligations?
The Parliament, backed by center-right MEPs from the European People's Party (EPP), pushed for a broad carve-out. Their argument: manufacturers of medical devices and industrial machinery already comply with extensive EU product-safety directives, and layering AI Act requirements on top creates duplicative regulatory burden. Germany's VDMA mechanical-engineering lobby made this case forcefully on April 27, warning that double regulation would slow innovation in industrial AI.
The Council, representing member states, showed what center-right MEPs described as "a lack of flexibility" on the exemption's scope. The Center for Democracy and Technology Europe noted in its April bulletin that the product-safety exemption "faced limited enthusiasm in the Council, with different compromise proposals being discussed." Center-left lawmakers, meanwhile, blamed the EPP and EU tech commissioner Henna Virkkunen for the breakdown.
Dutch lawmaker Kim van Sparrentak offered perhaps the sharpest reaction: "Big Tech is probably popping champagne. While European companies that care about safety and did their homework now face regulatory chaos."
Why the Calendar Matters More Than the Politics
The urgency behind these negotiations is not political — it is structural. The AI Act's core obligations for high-risk AI systems are set to apply from August 2, 2026, just three months away. The entire purpose of the AI Omnibus is to postpone that deadline: to December 2, 2027 for stand-alone high-risk AI systems, and to August 2, 2028 for those embedded in regulated products.
For that postponement to take legal effect before August, a final political agreement, formal Parliament vote, Council endorsement, and publication in the Official Journal must all occur in a matter of weeks. The legislative pipeline is unforgiving. If talks continue to stall in May and no agreement is reached before June, the original August 2026 deadline will stand. Full stop.
Adding to the pressure: the technical standards companies need to demonstrate compliance are not ready. Communications from CEN-CENELEC's Joint Technical Committee 21, which is drafting the harmonized standards, suggest the full set may not be available before December 2026, according to a client note from law firm Morrison Foerster.
The Industry-Rights Group Split
The failed trilogue has intensified an already sharp divide between industry groups and civil society organizations. CCIA Europe backed the December 2027 and August 2028 delays in March, citing gaps in technical standards and compliance guidance. VDMA pushed for the AI Omnibus to prevent double regulation for machinery.
On the other side, a coalition of 40 consumer, medical, and academic organizations — including BEUC, Access Now, EDRi, AlgorithmWatch, and Amnesty International — warned in an open letter this month that the Omnibus proposals "still risk reopening core elements of this framework, crucially weakening the AI Act." They described the changes as far more than a technical rewrite, arguing the package would leave EU citizens with less timely protection from high-risk uses such as biometric identification and AI systems deployed in schools.
What Companies Should Do Now
The practical advice for companies caught in this regulatory crossfire is straightforward, if uncomfortable. Enza Iannopollo, vice president and principal analyst at Forrester, offered a candid assessment of the enforcement landscape: "It's obvious that if the authorities responsible for enforcing the rules are not in place, there won't be enforcement, despite the deadlines. But Member States can accelerate that process and put those authorities in place rather quickly. Some countries have already named them. The risk is that businesses lose track of developments across each Member State and find themselves exposed to regulatory scrutiny and fines."
Companies should treat August 2 as a hard deadline regardless of what happens in May. If it gets delayed, that is a bonus. If not, it becomes an immediate regulatory risk for any firm deploying high-risk AI systems in the European market — particularly those in healthcare, automotive, critical infrastructure, and financial services.
Other parts of the AI Act continue moving on their original schedule. The prohibitions on unacceptable-risk AI have applied since February 2025. General-purpose AI rules came into force in August 2025. Transparency obligations under Article 50, including disclosure requirements for chatbot interactions and labeling of deepfakes, are set to apply from August 2.
What to Watch
The next trilogue is scheduled for May 13. If negotiators cannot close the product-safety exemption gap and reach a political deal before June, the original August 2026 deadline for high-risk AI obligations will apply as written. That outcome would force a compliance scramble across multiple industries — without the harmonized standards or national enforcement infrastructure that regulators themselves had assumed would be in place. The clock has not stopped ticking.
"If the authorities responsible for enforcing the rules are not in place, there will not be enforcement, despite the deadlines."— Enza Iannopollo, VP and Principal Analyst, Forrester