On July 31, 2026, the European Commission announced that from August 2, 2026 its AI Office, together with national authorities, will begin enforcing the Artificial Intelligence Act, Regulation (EU) 2024/1689. On the same date, Article 50 transparency obligations enter into application across the EU: interactive AI systems must disclose that users are dealing with AI, deepfakes must be labelled, and AI-generated or altered content must carry machine-readable marks.
The Commission announcement of July 31, 2026 and the companion press release IP/26/1714 mark the shift from soft-law guidance to live enforcement. Providers and deployers of chatbots, generative AI and synthetic media serving EU users must be operationally ready by August 2, regardless of the Digital Omnibus deferral of most high-risk Annex III obligations.
What starts on August 2, 2026 that did not apply before?
Two things land together. First, the Commission and Member State authorities open active AI Act enforcement for the layers already in force: prohibited practices, GPAI model duties where applicable, AI literacy, and now Article 50 transparency. Second, Article 50 itself becomes applicable: systems designed for direct interaction with natural persons must inform users they are interacting with AI; deepfake images, video and audio must be labelled; AI-generated or manipulated content must carry machine-readable marks so it can be detected.
The Commission also published a first list of more than 180 organisations that have signed the Code of Practice on transparency of AI-generated content, the endorsed operational route for marking and labelling compliance. Parallel complaint and whistleblower channels are live: the AI Act complaints tool, the AI Act whistleblower tool, and a channel for downstream providers using general-purpose AI models.
| Layer | Status from August 2, 2026 | Note |
|---|---|---|
| Article 50 transparency (disclosure, deepfake labels, machine-readable marks) | Applies | Unaffected by Digital Omnibus high-risk deferral |
| Prohibited AI practices / GPAI duties already in force | Under active enforcement | AI Office + national authorities |
| Annex III stand-alone high-risk obligations | Deferred (Digital Omnibus) | Planned to December 2, 2027 if Omnibus timing holds |
| Code of Practice on AI-generated content transparency | 180+ signatories published | Endorsed compliance pathway for marking/labelling |
Who must act by August 2, and who can still wait on high-risk rules?
Providers of chatbots and other interactive AI systems placed on the EU market or put into service for natural persons must ship clear AI-interaction disclosure. Providers of generative systems that output audio, image, video or text must implement effective, machine-readable marks. Deployers must label deepfakes and, where Article 50(4) applies, AI-generated or manipulated public-interest text that has not undergone genuine human review or editorial control. Third-country providers remain in scope when their system output is used in the EU.
Teams that had banked on the Digital Omnibus to push "everything AI Act" past August 2026 must separate calendars: high-risk Annex III conformity can wait under the Omnibus timeline; Article 50 and the already-applicable layers cannot. National market surveillance authorities will lead most Article 50 cases; the AI Office covers a narrower set of GPAI-linked and institutional situations.
What should compliance teams finish before the enforcement window opens?
Inventory every customer-facing chatbot, agent, avatar and generative pipeline that reaches EU users. Confirm disclosure copy and placement for Article 50(1), and select either Code of Practice adherence or a documented equivalent marking stack that is effective, reliable, robust and interoperable for Article 50(2). For deployers, lock human-readable deepfake labels and define when public-interest text has real editorial control versus token review. Brief product, trust and safety, legal and communications owners that complaints can now land through the Commission's published channels from day one.
Continuous, per-jurisdiction real-time monitoring surfaces Commission enforcement signals and Member State follow-on measures the moment they publish, so Article 50 readiness is not discovered after the first complaint.
Take advantage of this real-time watch
Confirm which of your AI products trigger Article 50(1) to (4) in the EU, lock August 2, 2026 into the compliance calendar separately from any high-risk Omnibus deferral, and decide within days whether to sign the transparency Code of Practice or document an equivalent programme. Register ownership for responding to AI Act complaints and whistleblower referrals. Obsidian keeps the EU AI Act enforcement track visible as national authorities and the AI Office begin active oversight.
