EU Starts Enforcing AI Act: Chatbots Must Now Identify as AI

My take: As of August 2, the EU AI Act has real enforcement power: any company using chatbots or generative AI with users in Europe must tell them they are talking to a machine, not a person. Deepfakes must carry a visible label, and AI-generated content must include machine-readable marks. Noncompliance can cost up to 15 million euros, or 3% of global annual revenue, whichever is higher.

This is not a law for the near future. It is in effect now, and it applies regardless of when a company launched its system. The only margin is for companies that already had systems on the market before August 2: they have until December 2 to meet the technical content-marking requirements.

For any business with a digital presence in Europe, the question is no longer "should we comply" but "are we ready to prove that we do." Transparency about what is AI and what is human becomes a legal obligation, not a design decision.

Does your team have documentation showing how every AI system it uses or deploys communicates its nature to users?