The EU AI Act rules that started on 2 August are the labelling ones, not the high risk ones

On 2 August 2026 the European Commission's AI Office and national authorities began enforcing parts of the AI Act, and the transparency obligations in Article 50 started to apply. The Commission's announcement is specific about what that means in practice: chatbots and other interactive AI systems must tell users they are dealing with AI and not a human, deepfakes must be labelled, and AI generated or altered content must carry machine readable marks so it can be detected more easily.

The shorthand that the AI Act came into force on that date is wrong in both directions. Parts of it were already enforceable. The prohibitions in Chapter 2 and the AI literacy requirements have applied since 2 February 2025. Obligations on general purpose AI models, the notified body framework, the governance chapter, confidentiality and the penalties in Articles 99 and 100 applied from 2 August 2025.

More importantly, the single biggest tranche of obligations that had been scheduled for 2 August 2026 did not arrive. The Digital Omnibus, which amends the AI Act, was approved by the European Parliament on 16 June 2026 and the Council on 29 June 2026, published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026, six days before the deadline it moved. DLA Piper's analysis records the effect plainly: the high risk compliance obligations that were due on 2 August 2026, including those covering recruitment, candidate selection, performance evaluation, task allocation, monitoring of workers and decisions on promotion or termination, now fall due on 2 December 2027.

Sources have not all caught up. The independent tracker artificialintelligenceact.eu still describes 2 August 2026 as the date on which the remainder of the AI Act applies, with only Article 6(1) held back to 2 August 2027, and contains no reference to the omnibus. A commentary published on 26 June 2026 by the data governance vendor Collibra described a provisional Digital Omnibus agreement reached in spring 2026 and said the deferrals were not yet final and should be confirmed, which was accurate when written, three days before the Council vote, and is no longer. An analysis by NicFab published on 30 July 2026 gives the post omnibus picture: Chapter IV transparency, Chapter IX market surveillance, the EU database, the AI Office enforcement powers in Articles 75a to 75d and the Article 101 penalties became operational on 2 August 2026, while Chapter III sections 1 to 3 did not.

What Article 50 requires is set out in the Commission's own FAQ. There are four duties, split between providers and deployers. Providers must ensure people know when they are interacting with an AI system, unless that is obvious, and must mark synthetic audio, image, video and text in a machine readable format so it is detectable as artificially generated or manipulated. Deployers must inform people exposed to emotion recognition or biometric categorisation systems, and must label deepfakes and AI generated text published on matters of public interest where there has been no human review or editorial control. Breaches carry fines of up to 15 million euros or 3 per cent of total worldwide turnover, with proportionality allowances for small and medium enterprises and small mid cap companies.

The timing is not quite as sharp as the enforcement date suggests. Systems placed on the market before 2 August 2026 have until 2 December 2026 to meet the marking obligation, and content generated before 2 August 2026 does not have to be labelled retroactively, though the Commission encourages it where possible.

The Code of Practice on transparency of AI generated content is the compliance route most organisations are expected to take. The Commission published the final version on 10 June 2026, after a process that opened with a consultation and call for participation in September 2025, and the Commission and the AI Board have confirmed the code as an adequate voluntary tool for demonstrating compliance. Signing is optional; the underlying Article 50 duties are not. Signatories commit to marking and labelling solutions that are effective, interoperable, robust and reliable as far as technically feasible.

The signatory count has been reported inconsistently by the Commission itself. Its 2 August announcement refers to a first list of more than 180 organisations that have signed. Its Code of Practice page states that around 190 companies and organisations had signed as of 31 July 2026. A public list of signatories has been published.

The Commission has also opened three reporting channels: an AI Act complaints tool, an AI Act whistleblower tool, and a separate complaints channel for downstream providers building on general purpose AI models.

The enforcement question is where the announcement is least informative. The AI Office does not have general authority over Article 50. Per the Commission's FAQ, its role is limited to general purpose AI systems where the same entity provides both the model and the system, and to systems integrated into designated very large online platforms. Everything else falls to national market surveillance authorities, and to the European Data Protection Supervisor where EU institutions, bodies and agencies are involved. Member states were required to designate national competent authorities and communicate them to the Commission by 2 August 2025.

What is not known is how many have done so competently, or at all. The Commission has not published an assessment of national readiness, and no Article 50 enforcement action has yet been made public. Until one is, the duty exists and the machinery for applying it is largely untested.