Starting August 2, if you chat with a bot anywhere in the EU, it must tell you "I'm AI" in its very first message.

The European Commission announced on July 31 that beginning August 2, the EU AI Office and competent authorities across all 27 member states would jointly enforce provisions of the AI Act. This marks the world's first comprehensive AI regulation moving from "legislative phase" to "enforcement phase."

What Takes Effect Today?

The first wave of mandatory enforcement centers on transparency provisions, directly affecting AI products that everyday users encounter:

First, all AI systems that interact directly with users — chatbots, AI customer service, virtual assistants, digital humans — must clearly inform users at the start of first interaction that "you are interacting with AI." No impersonating humans, no deliberately hiding AI identity during conversation.

Second, AI-generated or edited deepfake images, videos, and audio must be labeled and carry machine-readable markers so platforms and regulators can automatically detect and trace them. That means AI-generated content on social media won't just get a visible "AI-generated" watermark — it needs embedded technical markers that machines can detect too.

Third, the Commission published the first batch of signatories to the Code of Practice on AI-generated content transparency — over 180 organizations including tech companies, content platforms, and industry groups, committing to implement transparency requirements and cooperate with enforcement.

How Big Is the Impact?

The EU AI Act officially entered into force on August 1, 2024, with a phased implementation schedule. After the 2024 ban on unacceptable-risk AI and 2025 high-risk AI provisions, August 2, 2026 is when most core provisions — especially transparency requirements and general-purpose AI rules — officially become enforceable.

For Chinese companies, this means any product serving EU users must comply with transparency rules. Cross-border e-commerce AI customer service, chat features in global apps, AI content tools targeting the European market — all need to add AI identity disclosure and content labeling into product flows. Non-compliant products face fines up to 6% of global turnover.

The deeper impact is at the industry-standard level. EU regulation has a well-documented "Brussels Effect" — EU standards tend to become global de facto standards. GDPR reshaped global privacy compliance; the EU AI Act will likely do the same for AI transparency. When a market of 27 developed economies requires AI to identify itself, other countries and platforms will likely follow with similar rules.

Enforcement challenges remain significant. How to technically implement AI content labeling, what machine-readable marker standards look like, how to regulate cross-border AI services — these details will be worked out through enforcement. But August 2 is already in the history books: from this day forward, AI is no longer lawless territory in one of the world's largest developed economies.

See you tomorrow.