Quirin Schlegel

Loading

The deadline that didn't move

The Omnibus deferred high-risk obligations by 17 to 24 months and left Article 50 untouched; the four-month marking transition is a grandfather clause for incumbents, so a new agent shipped in August carries the full obligation from day one.

30 July 20269 verified claims9 sources

Everyone heard that the EU delayed the AI Act. Almost nobody checked which parts.

On Sunday, 2 August, Article 50 lands. It was never deferred — and it does not care whether your system is high-risk.

The Digital Omnibus on AI, Regulation (EU) 2026/1744, entered into force on 27 July. It pushed stand-alone high-risk obligations to December 2027 and product-embedded ones to August 2028. Article 50 transparency it left exactly where it was.

Four duties. Three of them touch any agent you have in production:

1. If your agent talks to a human, tell them it is a machine — at the latest at the first interaction. Not in the FAQ. Not in the terms. 2. If it generates audio, image, video OR TEXT, the output must be marked in a machine-readable format. Text is in scope. Most teams read "synthetic media" and mentally skip the line. 3. If that output goes out as a deepfake, or as text published to inform the public on a matter of public interest, that is a separate disclosure — and it sits with the deployer, not with your model vendor.

Here is the part I nearly got wrong. There IS a transition on the marking duty: recital 38 grants four months, to 2 December. But read who it is for — "providers who have already placed their systems on the market before the 2 August 2026". That is a grandfather clause, not a grace period. Ship a new agent on 3 August and you mark from day one. The transition rewards the incumbent and penalises the launch.

What I would do Monday:

1. Map every egress your agent has. Surfaces that reach a human need disclosure; surfaces that reach another system need marking. They are different obligations and they fail differently. 2. Put the marking in the emit path, not the UI layer. A badge in your frontend is not machine-readable. 3. Read the Code of Practice on Transparency of AI-Generated Content — final version 10 June, voluntary. Signing buys you one uniform compliance story instead of arguing equivalence with 27 national authorities.

€15M or 3% of worldwide turnover is the ceiling. The real cost is retrofitting provenance into a system that never had it.

Which of your agents can currently say "I am not a human" — in a form a machine can read?

#EUAIAct #AgenticAI #AICompliance #AIEngineering #Berlin

Evidence

What this rests on

Every factual claim in the edition above, stated on its own and checked before publication. If one of these is wrong, the argument is wrong — so they are listed where you can check them.

The European Commission adopted its guidelines on the implementation of the transparency obligations under Article 50 of the AI Act on 20 July 2026, together with a Code of Practice on Transparency of AI-Generated Content — thirteen days before those obligations start applying on 2 August 2026.
Why it mattersSets the lead time. Thirteen days is the whole argument for building to the article text rather than waiting for guidance, and it is the pattern to expect for the 2027 and 2028 high-risk deadlines too.
europeancom2026guideline · europeancom2026press
The Article 50 machine-readable marking obligation does not apply to source code, to short sequences of numbers, symbols or letters, to machine-to-machine outputs without human exposure, to outputs used only in closed-loop industrial and product development environments, or to systems performing an assistive function for standard editing.
Why it mattersThis is the carve-out that decides whether a coding agent is in scope at all. The code an agent emits is exempt from marking; almost nothing else the product does is. It reframes the compliance question from 'do we watermark generated code' to 'which of our surfaces did we forget to classify'.
europeancom2026faq
Article 50(1) requires providers of AI systems that interact directly with natural persons to make the AI nature known, where the interaction is a genuine two-way exchange. The exception for cases where it is obvious the person is interacting with an AI is to be interpreted in a restrictive manner, assessed against a reasonably well-informed, circumspect and observant average person.
Why it mattersMost teams assume a developer audience makes the AI nature obvious and skip the disclosure. The Commission says that exception is to be read restrictively and judged against an objective average person, not against your user base.
europeancom2026faq · stibbe2026transpare
Under a grandfathering rule introduced by the Digital Omnibus, generative AI systems placed on the market or put into service before 2 August 2026 must comply with the marking requirements only as of 2 December 2026. The concession covers marking and detection, not the disclosure duties.
Why it mattersThe only real relief available this year, and it is narrower than it sounds: it buys four months on watermarking and nothing at all on telling users they are talking to an AI.
lathamandwa2026ai · europeancom2026faq · stibbe2026transpare
Article 99(4) sets administrative fines for non-compliance with Article 50 at up to EUR 15 000 000 or, for an undertaking, up to 3 % of total worldwide annual turnover for the preceding financial year. Article 99(6) provides that for SMEs, including start-ups, the fine is capped at the lower of the percentage or the amount.
Why it mattersThe SME inversion in paragraph 6 is almost never quoted and it materially changes the exposure calculation for a small Berlin shop: the cap is the lower of the two figures, not the higher.
euartificia2026article · stibbe2026transpare
The Code of Practice on Transparency of AI-Generated Content was published on 10 June 2026 and is voluntary. It is undergoing an adequacy assessment by the Commission and the AI Board. The Commission states that following a positive assessment, signatories can rely on its measures to demonstrate compliance with the labelling and detection rules for AI-generated content, deep fakes and certain text publications. It covers Article 50(2), (4) and (5).
Why it mattersSigning today is a bet on the assessment landing positively, not a shield that exists now. The distinction decides whether the Code can be cited in a compliance file this quarter or only later.
europeancom2026code · stibbe2026transpare · europeancom2026press
The Digital Omnibus deferred EU AI Act obligations for stand-alone high-risk AI systems to 2 December 2027.
Why it mattersThe headline number in every EU compliance conversation for the next eighteen months.
lathamandwa2026ai
Obligations for high-risk AI in regulated products and safety components were deferred to 2 August 2028.
Why it mattersRelevant to anyone embedding agents in regulated products — including building-sector software.
lathamandwa2026ai
AI Act transparency obligations still apply from 2 August 2026 and were not deferred — covering interactive systems, emotion recognition, and biometric categorisation.
Why it mattersThis is the deadline that actually binds this year, and the one most teams have stopped tracking because they heard 'deferred'.
lathamandwa2026ai

Sources

AI Act Blogsecondary
AI Act Blog - Article 50 deadline
Original Archived PDFretrieved 30 July 2026 · published n.d.
EU AI Act Explorersecondary
Article 50- Transparency Obligations for Providers and Deployers of Ce
Original Archived PDFretrieved 15 October 2026 · published 1 January 2026
EU AI Act Explorersecondary
Penalties - Article 99
Original Archived PDFretrieved 12 October 2026 · published 1 January 2026
EUR-Lexprimary source
Regulation (EU) 2026-1744 (Digital Omnibus on AI)
Original Archived PDFretrieved 16 October 2026 · published 27 July 2026
Hunton Andrews Kurthsecondary
EU Digital Omnibus on AI Enters Into Force
Original Archived PDFretrieved 5 October 2026 · published 28 July 2026
Lewis Silkinsecondary
Lewis Silkin - AI labelling rules
Original Archived PDFretrieved 30 July 2026 · published 24 July 2026
Orricksecondary
Orrick - Digital Omnibus 8 changes
Original Archived PDFretrieved 30 July 2026 · published n.d.
European Commissionprimary source
Code of Practice on Transparency of AI-generated Content
Original Archived PDFretrieved 19 August 2026 · published 10 June 2026
Latham and Watkinssecondary
lathamandwa2026ai
Original Archived PDFretrieved 30 July 2026

Every source for this edition, as one PDF BibTeX — all editions

Corrections

What changed after publication

- **2026-07-30** - the draft described the four-month marking window as a grace period for everyone. Recital 38 of Regulation (EU) 2026/1744 limits it to providers who placed systems on the market before 2 August 2026. Corrected before publication; the corrected reading became the spine of the edition. - **2026-07-30** - detailed label placement rules (icon at the top of text, repeated for video, audible disclaimer for audio) were cut. Sourced to one law-firm briefing and not corroborated by the Commission's own page on the Code. Held at single-source in CLAIM-2026-07-30-07.