New EU AI rules mean your chatbot has to admit it is a chatbot
A fresh set of transparency requirements under the EU AI Act took effect this month, and they catch far more startups than the headline rules for high-risk systems ever did.
While the EU's toughest AI Act requirements for high-risk systems were pushed back to 2027, a separate, narrower set of rules quietly became enforceable this month, and it applies to a much wider range of everyday products.
Under the new transparency obligations, any chatbot or interactive AI tool has to tell users they are talking to a machine, not a person. AI-generated images, audio and video need machine-readable markers so they can be identified as synthetic, and deepfakes must be labeled outright.
The scope is broader than founders might assume. A startup does not need to be training its own model to be covered, a customer support bot or an AI-written marketing blurb served to EU users can trigger the obligation.
Penalties are real
Noncompliance can bring fines up to fifteen million euros or three percent of global revenue, whichever is larger. Surveys cited in the coverage suggest most small AI companies still have not finished classifying which of their systems the rule even touches.