The EU AI Act's enforcement regime goes live in 18 days — and the specific obligations that take effect August 2 are not the ones the European Parliament extended in May.
What Actually Happened
In May 2026, EU institutions reached a political agreement under the Digital Omnibus package that extended high-risk AI compliance obligations by sixteen months, pushing the deadline for standalone high-risk AI systems from August 2026 to December 2027. That headline was accurate and meaningful. But it covered a specific subset of the AI Act — the Annex III high-risk system requirements — and was widely interpreted as a broader stay of execution for the entire regulation.
It was not. Article 50, which governs transparency obligations for chatbots, AI-generated content, and deepfakes, takes effect August 2 on schedule. The European Commission's enforcement powers over general-purpose AI (GPAI) model providers — including the ability to request documentation, issue fines, and restrict market access — also activate on the same date. According to the EU AI Act's Article 101, fines for non-compliance reach up to 3% of global annual turnover or €15 million, whichever is higher.
There was a separate window that has already closed. The EU AI Office offered companies a path to presumed regulatory conformity by signing a Code of Practice on AI-Generated Content Transparency. The deadline to sign was July 22 — eight days ago. Companies that signed receive a presumption that they comply with their Article 50(2) and Article 50(4) obligations, reversing the evidentiary burden in any enforcement action. Companies that did not sign face the standard compliance burden without that presumption.
What This Means for the August 2 Deadline
Article 50 creates four binding obligations for any AI system deployed to EU users or interacting with EU nationals.
First, any AI system that interacts directly with people — a chatbot, a virtual assistant, an AI customer service interface — must disclose its AI nature before or at the start of the interaction. This disclosure must be clear and accessible.
Second, providers of generative AI systems producing text, images, audio, or video must mark their outputs in a machine-readable format that enables detection as AI-generated content.
Third, deployers using AI to create or manipulate audio or video content constituting a deepfake must label that content as artificially generated or manipulated.
Fourth, deployers using AI to generate text intended to inform the public on matters of public interest must disclose that the text is AI-generated.
These requirements apply to providers and deployers differently. An enterprise using a third-party chatbot is a deployer — but the disclosure obligation can fall on either party depending on the contractual arrangement. This is where many enterprises currently have a gap: their AI vendor contracts do not specify who is responsible for Article 50 compliance.
The Enterprise Lens
If your business uses a chatbot, AI-generated messaging, or any AI system that interacts with customers or the public in the EU — whether you built it or bought it from a vendor — the August 2 deadline is yours to manage.
The practical checklist is not long, but each item requires action your technology or legal team needs to take now. Any customer-facing AI interface that speaks, writes, or responds on behalf of your business needs a disclosure statement at the point of first contact. Any content your organisation generates using AI tools — articles, product descriptions, images, video — that is published to EU users needs to be labelled or marked in a machine-readable format. And every AI vendor contract you currently hold should be reviewed to confirm who carries the Article 50 compliance obligation, and whether your vendor has the documentation to prove they meet their end of it.
The question to ask your technology or legal team this week is not whether you are covered by the AI Act — if you have EU-facing AI products, you almost certainly are — but which specific systems trigger Article 50 obligations and whether the disclosures are already in place.
What to Watch
- Whether national market surveillance authorities in larger EU markets — Germany, France, the Netherlands — issue enforcement guidance or early actions in August, which would signal the pace and focus of the enforcement regime
- How GPAI model providers respond to European Commission documentation requests under their new enforcement powers, which will clarify what regulators actually expect in practice
- Whether the Article 50 Code of Practice becomes a de facto compliance standard over time — even for companies that missed the July 22 signing window, the document describes the practices regulators will expect to see
Sources
- EU AI Act Article 50 Transparency Rules Guide
- EU AI Act Enforcement Is Here: Chatbot Rules Live, High-Risk AI Delay Now Binding Law
- EU AI Act Chatbot Disclosure and Deepfake Labeling: July 22 Signatory Deadline
- Morgan Lewis: EU Approves Delays and Amendments to Certain EU AI Act Obligations
- GPAI Enforcement and 3% Fines Begin August 2