If your SaaS product uses AI and serves EU customers, the Act applies, even if you're headquartered outside Europe. Most AI features trigger at least Article 50 transparency, in force since August 2, 2026. Some are high-risk, with a December 2, 2027 deadline.
Article 50 says people interacting with AI should know it. That covers nearly every AI-powered feature.
Any AI chatbot, virtual assistant, or support agent that interacts with EU users must disclose it is AI-powered.
Text generation, code completion, email drafting, or any feature where AI creates content users may believe is human-written.
Synthetic media features, avatar generators, or AI-created visuals must be labelled as AI-generated.
AI that analyses tone, emotion, or intent in user communications must inform users the analysis is happening.
If your AI makes or influences decisions about people in these categories, the full set of high-risk obligations applies.
Not everything with "AI" in the name triggers obligations. Rule-based systems and internal-only tools often sit outside it.
Not sure? The free classifier tells you exactly where your system falls.
If you build and ship the AI model, you're a provider, and the full obligations apply. If you integrate a third-party AI (OpenAI, Anthropic, and the like) into your product, you may be a deployer with lighter obligations, but you're still responsible for transparency, human oversight, and using the system as intended. Many SaaS companies are both: provider of their product, deployer of the underlying model.
Read: Deployer vs Provider, which are you?The classifier is free, no account required. Answer six questions and get your risk tier, the applicable articles, and a clear next-steps checklist.
Classify your AI system