Four months. That’s the gap compliance teams may have missed.
The Annex III headline, high-risk AI systems now face December 2, 2027 instead of August 2, 2026, has dominated post-Omnibus coverage. Sixteen months of runway is a number that registers. Two quieter Omnibus provisions deserve equal attention for different audiences.
According to Hogan Lovells’ analysis of the Digital Omnibus on AI, Article 50(2)’s synthetic content marking requirement has been delayed four months for a specific set of AI systems. The new deadline is December 2, 2026, moved from August 2, 2026. The catch, and this scoping condition matters legally, is that the delay applies only to systems already deployed before August 2, 2026. Systems that launch on or after August 2 face the original obligation structure under the Act.
This isn’t a blanket grace period for GenAI providers. It’s a targeted reprieve for legacy deployments.
Definition
Article 50(2) requires providers of AI systems that generate or manipulate synthetic content, text, images, audio, video, to ensure that content is marked in a machine-readable format. The EU hasn’t finalized approved technical standards for that marking as of this reporting. That gap between a December 2026 compliance deadline and the absence of finalized standards is the real calendar problem. Providers of already-deployed systems have four extra months, but the standards they’re supposed to implement may not be settled before the new deadline arrives.
The second under-reported change: the requirement for each EU member state to establish at least one AI regulatory sandbox has been deferred from August 2, 2026 to August 2, 2027, per Hogan Lovells’ reading of the Omnibus. That’s a 12-month deferral for member state governments, not for industry, but it has downstream implications for companies that were planning to use sandbox access to test compliant deployments ahead of high-risk deadlines.
What doesn’t change: the prohibited AI practices provisions, in force since February 2025, are untouched. GPAI model transparency obligations, effective August 2025, are untouched. Those two anchors of the EU AI Act’s enforcement architecture remain fully operative.
The full revised deadline picture, per Hogan Lovells’ analysis: Article 50(2) synthetic content marking (pre-August deployed systems), December 2, 2026. Annex III high-risk systems, December 2, 2027. Annex I product-embedded AI, August 2, 2028. National sandboxes (member state obligation), August 2, 2027.
Unanswered Questions
- What technical standards will satisfy Article 50(2)'s machine-readable marking requirement, and will they be finalized before December 2, 2026?
- Does the August 2, 2026 deployment cutoff apply to systems that have been updated or fine-tuned after that date, or only to systems with no changes?
- Which national sandbox programs are planned and on what timeline, given the 12-month member state deferral?
Don’t expect the standards problem to resolve itself. The Article 50(2) deadline moved, but the underlying technical standards gap didn’t. GenAI providers with systems deployed before August 2 should use the four-month window to push EU standards bodies for timeline clarity, not just update their compliance calendars and assume the standards will arrive in time.
All deadline information in this brief is sourced to Hogan Lovells’ legal interpretation of the Omnibus text. The EU Commission’s primary publication hasn’t been directly cited in this item, operators should verify against the official text before updating compliance programs.