August 2 is circled on your calendar. Here's what actually switches on.

Your inbox is full of "August 2" warnings. Some are accurate. Some are stale. You need the real picture before you panic or ignore it.

The quick version for you

August 2, 2026: Article 50 transparency rules become fully enforceable. EU regulators can actively pursue you — not just publish guidance.

Most high-risk AI system rules got pushed to December 2, 2027. That happened under the EU Digital Omnibus deal (May 7, 2026). Article 50 kept its original date.

What Article 50 requires from you starting August 2

Three things become enforceable:

1. AI chatbot disclosure You deploy AI that talks directly to people? Tell them it's AI. Before or at the start of the interaction. Example: your Drift bot on the pricing page. Add "This chat uses AI" to the welcome message.

2. AI-generated content labelling (December 2, 2026 — slightly later) AI-generated or manipulated images, audio, and video need machine-readable AI labels. This targets deepfakes and synthetic media. Text gets a later deadline. Example: Midjourney hero images on your landing page.

3. Emotion recognition and biometric categorisation disclosure Your system detects emotions or sorts people by biometric traits? Tell affected individuals. Rare for most small teams. Example: an HR tool that scores interview "confidence" from video.

What doesn't change on August 2

The big delay: Annex III high-risk AI obligations. These cover AI in employment, education, credit scoring, healthcare, law enforcement, and critical infrastructure. Original August 2, 2026 deadline? Moved to December 2, 2027.

If you thought high-risk rules hit August 2, you likely have more time.

The delay doesn't touch Article 50. It doesn't touch Article 4 either. Article 4 (AI literacy) has been live since February 2025. You're already on the clock for that one.

What to do before August 2

You have a customer-facing chatbot or AI assistant: Add a disclosure today. "This service uses AI" at the top of the chat works. Get it live before August 2.

Your team uses AI tools (ChatGPT, Copilot, Notion AI): Run AI literacy training. Document it. Article 4 has been enforceable since February 2025. If you haven't trained your team yet, you're already behind.

You're not sure August 2 applies to you: Run the free NudgeCompliant audit. Four minutes. It tells you exactly what hits your setup.

A word on enforcement

Fines can reach €35M or 7% of global turnover. For small teams, enforcement should stay proportionate. Regulators won't hunt a 20-person agency on day one for a missing chatbot line.

"Proportionate" isn't "non-existent" though. The bigger risk for UK teams isn't fines. It's procurement. Enterprise clients ask for AI compliance docs before they sign. Missing a disclosure can cost you a contract.

August 2 is a deadline. It's also a sales edge. Sorted compliance beats competitors who haven't started.

Your 10-minute action: List every place EU customers interact with your AI — chat widgets, auto-emails, support bots. Put a checkmark next to each one that already says "This uses AI." Fix the gaps today.

Article 50 obligations active August 2, 2026. Information accurate as of [date]. Not legal advice.