The window is closed.
As of today, June 3, 2026, the European Commission’s consultation on Article 50 draft transparency guidelines has concluded. The comment period on how AI systems must label and watermark synthetic content is over. What the Commission does with those submissions, and how quickly, determines whether practitioners have any adjustments to plan for before August 2.
That’s the deadline that wasn’t deferred.
The Digital Omnibus provisional agreement, reached on May 7, 2026, pushed several EU AI Act compliance dates significantly. But Article 50’s transparency obligation for newly deployed AI systems was not among them. August 2, 2026, remains active. Any AI system deployed after that date must comply with the Article 50 transparency requirements as they stand, including synthetic content labeling and, for applicable systems, watermarking. Compliance teams that absorbed the Omnibus deferral news and assumed broad deadline relief need to check their August 2 exposure now.
The three deadlines that matter between today and December 2.
June 23 is next. The consultation on draft guidelines for classifying high-risk AI systems under Article 6 and Annex III closes in 20 days. This is the consultation that determines how “high-risk” gets defined in practice, which specific systems fall under the Annex III list and what evidence is required to establish or dispute that classification. Prior coverage confirmed this deadline across multiple pipeline cycles. If your organization is assessing whether a deployed system qualifies as high-risk, June 23 is the last point at which you can formally influence the classification guidance.
August 2, 2026 arrives next. Article 50 transparency obligations for newly deployed AI systems take effect. This deadline was not moved by the Omnibus. Providers of AI systems in the EU, including synthetic image generators, voice cloning tools, and deepfake detection systems, must have labeling and watermarking processes in place by this date.
December 2, 2026 is the final marker in this window. Article 5’s new prohibitions take effect, including the absolute ban on AI-generated non-consensual intimate imagery (“nudifiers”) and child sexual abuse material. These weren’t deferred, they were added by the Omnibus as strengthened protections with a December 2 effective date. On the same day, Article 50(2) synthetic content marking obligations for legacy systems (those deployed before August 2) also take effect, per the Omnibus deferral provisions.
What the Omnibus deferred, and what it didn’t.
The distinction matters for planning. Annex III stand-alone high-risk AI systems – covering employment screening, credit scoring, and law enforcement applications – have their compliance deadline pushed from August 2, 2026, to December 2, 2027. That’s a 16-month extension, confirmed by DLA Piper’s client alert. Annex I embedded high-risk AI, medical devices, aviation hardware, is reported to be deferred to August 2, 2028, according to compliance analyses of the Digital Omnibus agreement.
But neither deferral touches Article 50 for newly deployed systems. That’s the practical compliance asymmetry the Omnibus created: significant relief for high-risk classification obligations, none for transparency obligations on new deployments.
What to watch. The Commission will now process the Article 50 consultation submissions and finalize the transparency guidelines. Watch for final guidance publication in July 2026, that’s the last realistic window to operationalize any changes before August 2. If guidance arrives after August 1, practitioners will be implementing requirements under provisional interpretations. That’s a manageable but imperfect compliance posture.
TJS synthesis. The consultation closing is an administrative milestone, not a deadline in itself. The real deadline sequence runs: June 23 (classification guidance), August 2 (Article 50 deployment), December 2 (Article 5 prohibitions and legacy watermarking). Don’t expect the Commission to issue final guidance with significant lead time, the EU AI Act implementation pattern has consistently compressed the gap between guidance publication and effective dates. Organizations that haven’t started Article 50 implementation work are already behind the realistic schedule, not just the regulatory one.