The EU Now Requires AI to Identify Itself
Since August 2, 2026, new transparency rules under the EU AI Act require AI systems to be upfront about what they are — chatbots have to tell you you're talking to an AI, and certain AI-generated content has to carry a clear label. Here's what it actually covers, and whether it affects you outside the EU.
⚡ Quick facts
- In effect since: August 2, 2026, under the EU AI Act
- Covers: Deepfakes, emotion-recognition tools, and unreviewed AI-written public content — plus any chatbot/avatar interaction
- Enforced by: National authorities, the European AI Office, and the EDPS
- Outside the EU: Not legally binding, but often rolls out globally anyway since most AI products aren't built as EU-only versions
Two separate requirements
The rules cover two distinct situations:
- Content labeling: certain AI-generated or manipulated content — deepfakes, emotion-recognition tools, and AI-written public-interest content published without human review — must be clearly and visibly labelled as such.
- Interaction disclosure: if you're talking to a chatbot or avatar instead of a real person, the system has to make that clear to you.
The EU has also published standardized icons that companies can use to mark AI-generated content consistently, along with compliance guidelines for organizations.
Who actually has to follow this
The obligation falls on providers and deployers of AI systems operating in the EU market. Enforcement is split across national market surveillance authorities, the European AI Office, and the European Data Protection Supervisor. These transparency rules are part of the broader EU AI Act, which originally entered into force back in August 2024 — this is one of its later provisions taking effect on schedule.
Does this matter if you're not in the EU?
Legally, the requirement only binds AI systems operating in the EU. In practice, though, most large AI companies build one product for a global user base rather than maintaining a separate EU-only compliant version — so features like chatbot disclosures or AI-content labels introduced for EU compliance often quietly show up for everyone, everywhere, over time. Worth keeping an eye on if a chatbot or AI tool you use suddenly starts labeling itself more clearly.
EU AI Act Risk-Tier Architecture & GPAI Technical Requirements
The European Union's AI Act establishes a horizontal regulatory matrix that categorizes AI systems into four risk tiers: Unacceptable Risk (prohibited outright), High Risk (strict pre-market conformity assessments), General-Purpose AI / GPAI with Systemic Risk (frontier compute models exceeding 10^25 FLOPs), and Minimal Risk. Transparency obligations mandate machine-readable provenance metadata, automated synthetic media labeling, and detailed energy consumption disclosures.
Frequently asked questions
What do the rules require?
Labeling certain AI-generated content, and disclosing when you're talking to an AI rather than a person.
When did this take effect?
August 2, 2026, under the EU AI Act.
Who must comply?
Providers and deployers of AI systems operating in the EU market.
Does it affect non-EU users?
Not legally, but global products often roll the same compliance features out everywhere.