The EU AI Act's August 2 Deadline Got Delayed. Not All of It.

If you spent the first half of 2026 bracing for the EU AI Act's next major deadline and then heard it got postponed, you have roughly half the story. On 29 June 2026, the Council of the EU gave final approval to the Digital Omnibus on AI — the first substantive amendment to the Act since it entered into force in 2024 — and pushed the high-risk system obligations that most compliance teams had been building toward back to 2 December 2027 for stand-alone Annex III systems, and 2 August 2028 for AI embedded in already-regulated products.

That is genuine, welcome breathing room for anyone building a high-risk system: a hiring tool, a credit-scoring model, a medical device with an AI component. It is also the reason a lot of businesses are walking into this weekend assuming 2 August 2026 no longer matters. It does. The Digital Omnibus deferred the high-risk track. It left Article 50 — the Act's transparency obligations — on its original schedule. If your business runs a customer-facing chatbot, generates marketing copy or images with AI, or publishes AI-synthesized content, Monday's deadline still applies to you, whether or not you touch anything the Act classifies as "high-risk."

We wrote in our July piece on Swiss AI regulation that the compliance work does not wait for the headline law to land. This is the same pattern one level up: the EU deadline that made headlines got real relief, and the deadline nobody was tracking is the one that actually bites this week.

What was pushed back, and by how much

The Digital Omnibus followed the European Commission's late-2025 proposal, a provisional agreement between Council and Parliament on 7 May 2026, Parliament's endorsement on 16 June, and final Council sign-off on 29 June. The result:

• High-risk AI systems under Annex III (recruitment tools, credit scoring, biometric categorisation, and similar) — obligations deferred to 2 December 2027.
• AI embedded in products already regulated under other EU law (medical devices, machinery, toys) — deferred to 2 August 2028, aligning with those products' own conformity cycles.

If your roadmap included a conformity assessment, technical documentation, or EU database registration for a high-risk system this August, that work now has more runway. It is worth reconfirming with counsel exactly which systems in your stack fall under Annex III before assuming the extension applies — the deferral is specific to the high-risk track, not a blanket pause on the Act.

What did not move: Article 50

Article 50 governs a narrower but far more common set of AI systems: anything that talks to a person, generates synthetic content, or produces a deepfake. None of that was part of the Digital Omnibus deferral, and it takes effect on 2 August 2026 as originally written. In practice, four obligations now apply:

Chatbots and AI agents must disclose that they are AI. Any system designed to interact directly with a natural person — a support chatbot, a sales assistant, a voice agent — must make it clear to that person that they are talking to an AI, unless it would already be obvious from the context to a reasonably well-informed user.

AI-generated or manipulated audio, image, or video must be labelled as a deepfake. A deployer producing synthetic media that could be mistaken for authentic content must disclose that it has been artificially generated or manipulated, at the latest upon the person's first exposure to it.

AI-generated text published to inform the public on matters of public interest must be disclosed as such, unless a human has reviewed it and takes editorial responsibility for the content.

Emotion-recognition and biometric-categorisation systems must inform the people they process. If a system infers emotional state or categorises someone by biometric data, the people it processes must be told.

The one piece of Article 50 that did get transitional relief is narrower than the disclosure duties themselves: the machine-readable watermarking and detection requirement under Article 50(2), which technically marks AI-generated content for automated detection rather than human-readable labelling. Providers of systems already on the market before August 2026 get until 2 December 2026 to implement that specific technical marking. The disclosure obligation — telling the person in front of the content that it is AI-generated — is not part of that carve-out.

Why this is not just an EU problem for Swiss companies

Switzerland has no AI-specific law in force, and its own consultation draft is not due until the end of 2026. That has led some Swiss SMEs to treat EU AI Act deadlines as background noise. Article 50 does not work that way: it applies based on where the output reaches a person, not where the provider is headquartered. A Swiss company running a customer-facing chatbot for EU customers, publishing AI-assisted content that reaches EU readers, or offering an AI tool into the EU market is a "provider" or "deployer" under the Act for that activity, regardless of where the company itself is based.

The enforcement architecture behind this is also live from 2 August 2026, independent of the high-risk deferral: each member state's designated market surveillance authority gains its investigation and enforcement powers on schedule, and the Act's tiered penalty structure — up to €35 million or 7% of global annual turnover for the most serious violations, with a lower cap that applies specifically to SMEs — applies to violations of the obligations that are actually in force, transparency included.

What to check before Monday

Audit anything customer-facing that uses AI. Chatbots, voice agents, and AI sales assistants need a disclosure a person will actually notice — not a line buried in a terms-of-service page nobody reads.

Check your content pipeline for synthetic media. If marketing, support, or product teams generate images, video, voiceovers, or AI-assisted articles that reach the public, confirm each one carries the disclosure Article 50 requires, and that editorially reviewed content is documented as such.

Do not assume the Digital Omnibus bought you time here. It bought real time on Annex III high-risk work. It bought nothing on chatbot disclosure, deepfake labelling, or AI-generated content notices — if your team paused that workstream when the delay headlines broke, restart it this week.

Treat this the way we described Swiss FADP compliance in July: the underlying obligation — tell people when they are dealing with AI — is the kind of thing worth building into a product from day one rather than retrofitting under deadline pressure. It is also, not coincidentally, good practice regardless of which jurisdiction's law is doing the asking.

The bottom line

The Digital Omnibus is genuinely good news for anyone with a high-risk AI system on their roadmap — two more years to get conformity assessment and documentation right is meaningful relief. But it is a deferral of one track of the AI Act, not a rewrite of the whole thing, and the transparency obligations that touch the most common AI use cases — a chatbot on your website, an AI-generated image in a campaign, a synthetic voice on a support line — were never part of that deferral. If your compliance checklist got shorter when the delay news broke, it is worth checking exactly which line items actually moved.

If you want a second opinion on whether your customer-facing AI systems meet Article 50's disclosure requirements before Monday, get in touch with our team.

Sources

• EU AI Act Omnibus Agreement — Postponed High-Risk Deadlines and Other Key Changes — Gibson Dunn
• EU agrees to delay key AI Act compliance deadlines — Travers Smith
• Article 50: Transparency Obligations for Providers and Deployers of Certain AI Systems — EU Artificial Intelligence Act
• The EU AI Act's Transparency Rules: A Practical Guide to Article 50 — EU Artificial Intelligence Act
• Deepfakes, Chatbots, AI-Generated Text: European Commission Details Transparency Obligations Under the AI Act — Greenberg Traurig
• Transparency obligations under Article 50 of the AI Act — European Commission
• Key Issue: Fines/Penalties — EU AI Act