Real numbers from this tool, line by line.
If your SaaS product ships an AI chat assistant, an image generator or a voice feature, the EU AI Act changed your screens on 2 August 2026. That is the date Article 50 started to apply (Article 113 of Regulation (EU) 2024/1689).
What Article 50 asks the screen to show
Article 50(1): people must be informed that they are interacting with an AI system, unless that is obvious to a reasonably well-informed person. In a SaaS app, that is a chatbot notice at the start of the conversation. Article 50(5) adds the timing: clear and distinguishable, at the latest at the first interaction or exposure. A notice buried in the terms of service does not meet that.
Article 50(2): providers of systems that generate synthetic audio, image, video or text must mark the output in a machine-readable format. That part lives in the file metadata, but the screen still needs a state that says the latent marking stays in place when a user turns the visible label off.
Article 50(3): deployers of emotion recognition or biometric categorisation must inform the people exposed to it.
Article 50(4): deepfakes must be disclosed as artificially generated or manipulated. AI text published to inform the public on matters of public interest must be disclosed too, unless it went through human review or editorial control and a person or company holds editorial responsibility. That exception is a checkbox and a named editor on your publishing screen.
What getting it wrong costs
Article 99(4) sets fines of up to €15,000,000 or 3% of total worldwide annual turnover, whichever is higher. For SMEs and start-ups, Article 99(6) flips it: whichever is lower.
We ran three cases through the free checker on the product page:
- A chatbot at €800,000,000 turnover, not an SME: 3% is €24,000,000, above €15,000,000, so the cap is €24,000,000.
- The same chatbot at €100,000,000 turnover: 3% is €3,000,000, so the cap stays at €15,000,000.
- An SME at €10,000,000 turnover: the lower amount applies, €300,000.
California is a different rule
California SB 942, the AI Transparency Act, covers GenAI providers with over 1,000,000 monthly users. It asks them to offer users a visible (manifest) disclosure on image, video and audio content and to embed a latent one. It does not cover chat text. The penalty is $5,000 per violation, and each day counts as a separate violation: an image generator that ran 30 days without the option comes to $150,000 in the checker. At 900,000 monthly users, the same feature has no SB 942 duty.
So one product needs two rule sets: a chatbot needs the EU notice but nothing under SB 942, and an image generator needs the export toggle for California and the machine-readable marking for the EU.
What the kit draws
The SaaS UI kit holds 8 Figma components with 27 variants. There is an AI content label with 4 variants, a label explainer tooltip with 3, and an AI interaction notice with 4 (including emotion recognition and a missing-label error). The visible-label export toggle has 3 variants and the publisher confirmation checkbox has 3. There are 3 AI output cards, 3 EU and California market tabs, and 4 disclosure actions. It ships as a Figma plugin, a .sketch file that Figma imports by drag and drop, and one SVG per variant. You swap the 6 color tokens for your brand.
The label and the tooltip are free: 7 variants, enough to label AI output on any page. The rest covers the screens around the label.
It is a design file, not legal advice: it puts the Article 50 and SB 942 states into your design review before the feature ships.