For months, August 2, 2026 has been the date. It’s the deadline on which EU AI Act obligations for high-risk AI systems were set to apply, the target that prior reporting on this hub told compliance teams to prepare against. Now, trilogue negotiations on the EU’s Digital Omnibus package are proposing to move that date, by as much as two years, depending on the system type.
Legal analysts tracking the negotiations report three proposed deadline tracks. For standalone high-risk AI systems, the Digital Omnibus reportedly proposes a new deadline of December 2, 2027. For high-risk AI embedded in regulated products, medical devices, machinery, vehicles, the proposed deadline extends further, to August 2, 2028. A separate proposed extension would push the watermarking compliance deadline to November 2, 2026. These figures come from law firm analysis by Jones Walker and A&O Shearman, not from official EU Commission documents, every proposed date must be treated as subject to change until formal adoption.
The A&O Shearman article title is itself instructive: “EU AI Omnibus: Key Issues as Trilogue Negotiations Begin.” Trilogue is the EU’s three-way negotiation between the Commission, Parliament, and Council. Nothing is final until all three agree on consolidated text and the formal adoption process completes. Legal analysts project a political agreement could be reached as early as April 28, 2026, though that projection hasn’t been confirmed and should be treated as an estimate.
The Digital Omnibus may also clarify the scope of high-risk classifications, potentially narrowing which AI applications face the full compliance burden. Legal analysts suggest a carve-out for convenience or optimization use cases is on the table, though the precise language remains in negotiation. For organizations whose systems sit near classification boundaries, this scope question matters as much as the deadline dates.
What the Digital Omnibus does not propose changing: the prohibitions on unacceptable-risk AI practices that have already taken effect, and the GPAI (general-purpose AI) obligations already in force. Compliance work on those tracks is unaffected by the proposed deadline shifts. Teams that have built documentation frameworks, conducted conformity assessments, or established AI governance structures haven’t wasted that work, much of it applies regardless of which deadline track ultimately governs.
The standard legal advice, echoed across the firms tracking these negotiations, is to continue planning against the original August 2, 2026 deadline. The reasoning is sound. Proposals in active trilogue can be modified, narrowed, or withdrawn. An organization that pauses compliance investment in anticipation of a two-year delay and then finds the proposals didn’t survive negotiation would face a compressed timeline with no good options. Maintaining August 2 as the working deadline is a hedge against that risk.
What to watch: The trilogue timeline itself, any political agreement announcement, the formal Commission publication of amended deadline language, and whether the high-risk classification scope clarification survives negotiation intact. The April 28 projected agreement date is the next near-term marker.
TJS synthesis: The Digital Omnibus, if adopted as proposed, would give compliance teams breathing room they didn’t have a month ago. But “proposed” and “adopted” are separated by a political negotiation that hasn’t concluded. The only defensible compliance posture right now is to treat the relief as a possible future state while continuing to build against the deadline that’s actually on the books.