The EU AI Act deadline that didn't move: what applies from 2 August

If you followed the headlines over the summer, you probably came away with one message about the EU AI Act: it's been delayed. That's true, and it's also the most dangerous half-truth in tech compliance right now.

In June, the European Parliament approved amendments pushing back the Act's high-risk system obligations: the heavyweight rules covering AI used in areas like recruitment, credit decisions and critical infrastructure. Those now arrive in December 2027 and August 2028, giving providers and deployers considerably more breathing room.

But the transparency duties in Article 50 were not delayed. They apply from 2 August 2026, and they're the part most likely to touch an ordinary business.

What actually applies from 2 August

Three duties, in plain terms:

• Tell people when they're talking to a machine. If you run a customer-facing chatbot, users must be clearly informed they're interacting with AI, in the conversation itself, not buried in your terms and conditions.
• Mark AI-generated content. Generative AI outputs (text, images, audio, video) need machine-readable marking identifying them as AI-generated. (Systems already on the market get a grace period on the technical watermarking, until 2 December 2026.)
• Label deepfakes. Synthetic content depicting real people or events that could pass as authentic must be disclosed as artificially generated.

The penalties are not token: breaches of these transparency duties can attract fines of up to €15 million or 3% of global annual turnover, whichever is higher.

"But we're a UK business"

The AI Act applies to businesses serving users in the EU, wherever the business is based. If your chatbot talks to EU customers, or your AI-generated marketing reaches an EU audience, the duties can reach you, Brexit notwithstanding. And even for purely domestic firms, this is the clear direction of travel: being upfront about AI use is rapidly becoming the baseline customers expect, regulation or not.

The good news: this is cheap to fix

Unlike the high-risk regime, which involves risk management systems, conformity assessments and technical documentation, the transparency duties are mostly a design exercise:

1. Audit where you use AI in anything customer-facing: chatbots, generated content, marketing imagery.
2. Add clear, in-context disclosure to conversational AI.
3. Make sure your content tools apply the appropriate marking to generated output.
4. Write down what you use AI for: future rules will reward businesses that already know.

We build AI features into business software with exactly these obligations in mind: disclosure and marking are far easier designed in than retrofitted. If you're not sure where your AI use stands, get in touch and we'll help you take stock.