AI Transparency Duty in 24 Hours: The Emergency Checklist for Swiss SMEs
In short
From 2 August 2026, the transparency obligations under Article 50 of the EU AI Act become enforceable — including for Swiss SMEs without an EU presence, once their AI outputs are used in the EU. Companies lacking chatbot disclosure, labelled AI content, AI literacy, and an AI inventory risk fines up to EUR 15 million or 3% of global turnover.
What Actually Happens on 2 August 2026
On 31 July 2026, the European Commission confirmed by press release what many Swiss companies have missed: as of tomorrow, 2 August 2026, the transparency obligations under Article 50 of the EU AI Act become enforceable. This applies to your company even without an EU establishment — as soon as your AI systems produce outputs used in the EU, the regulation applies.
The Dangerous Misconception
Many executives believe the Digital Omnibus of May 2026 postponed every AI Act deadline. That is only true for high-risk obligations under Annex III, now pushed to December 2027 and August 2028. Article 50's transparency duties were never part of that delay — they take effect tomorrow, regardless of company size.
EUR 15 million or 3%
Fines under Article 99 can be substantial; for many breaches they are set at up to EUR 15 million or 3% of global annual turnover.
What Article 50 Actually Requires
Article 50 covers four situations in which users must be able to recognise that they are dealing with artificial intelligence. The obligation applies regardless of whether you built the AI application yourself or purchased it from a vendor.
- Chatbots and voice assistants: every interaction with an AI system must be recognisable to the user — unless it is already obvious from the context.
- AI-generated content: text, images, audio, and video that were generated or substantially altered by AI must be labelled as such.
- Deepfakes: image, audio, or video material that convincingly imitates real people, objects, or events requires explicit disclosure.
- Emotion recognition and biometric categorisation: individuals must be informed when a system is used to detect their emotions or categorise them biometrically.
The 24-Hour Checklist: 4 Measures to Implement Right Now
The following four steps can be implemented within a day, without launching a full compliance project. They close the most urgent gaps until a complete AI Act compliance structure is in place.
1. Add Chatbot Disclaimers
Add a visible notice to every chatbot, virtual assistant, and AI-powered customer service tool. A proven wording is: 'You are communicating with an AI system. For questions, contact [contact].' In most cases, a text snippet in the first chat window or a meta tag in the widget header is enough — implementation often takes only a few hours.
2. Label AI-Generated Content
Review every channel where AI-generated text, images, or video is published — website, social media, newsletters, product descriptions. A simple addition such as 'Created with AI assistance' in the footer or caption is sufficient in most cases. Watermarking for generative tools already on the market before 2 August 2026 carries a grace period until 2 December 2026 — tools introduced after that date must comply immediately.
3. Secure AI Literacy Across Your Team
Article 4 requires an adequate level of AI literacy for everyone who operates AI systems or uses their outputs. A short internal briefing covering the essentials — when disclosure is required, who is responsible internally — is sufficient as an immediate measure.
4. Build an AI Inventory
List every AI application used across your company in a simple table: purpose, provider, data flow, and affected groups. This inventory forms the foundation for every further compliance step you will need to take in the coming weeks.
Does This Apply to Swiss Companies Without an EU Presence?
Yes. The EU AI Act has extraterritorial reach: as soon as the outputs of your AI systems — an AI-generated ad, a chatbot conversation, an automatically produced image — are used within the EU, your company falls under the corresponding obligations. For a Swiss SME with customers or subsidiaries in the EU, this is not a theoretical question; it is enforceable law starting tomorrow.
AI Compliance Is a Board-Level Responsibility, Not an IT Project
Delegating responsibility for AI transparency to the IT department underestimates the issue. It concerns legal risk, brand reputation, and customer trust simultaneously — decisions that belong at the executive level. The Swiss AI podcast on making AI a leadership priority explores how mid-market companies can anchor this principle in practice.
In international comparison, Switzerland leads Germany in AI adoption but trails the US and China — a gap that can widen further if executives freeze in the face of regulatory uncertainty instead of acting. The Swiss AI podcast on AI strategy for SMEs in the global race outlines how to avoid that paralysis.
What KI-Outsourcing.ch Handles for You
As an external AI division, we take on exactly these tasks for Swiss companies: from the immediate checklist through the complete AI inventory to ongoing monitoring of new regulatory deadlines. That keeps AI compliance a managed function rather than a recurring emergency.
Frequently asked questions
- What fines apply for non-compliance with Article 50?
- Up to EUR 15 million or 3% of global annual turnover, whichever amount is higher, under Article 99 of the EU AI Act.
- Were AI Act deadlines postponed by the Digital Omnibus?
- Only partially. The May 2026 Omnibus postpones high-risk obligations under Annex III to December 2027 and August 2028. Article 50's transparency obligations were not affected and apply from 2 August 2026.
- Does the transparency duty apply to Swiss companies without an EU establishment?
- Yes, provided the outputs of their AI systems are used within the EU (Article 2(1)(c) of the EU AI Act).
- What counts as AI-generated content that must be labelled?
- Text, images, audio, and video that were generated or substantially altered by an AI system, including deepfakes.
- Is there a grace period for watermarking generative AI tools?
- Yes, tools already on the market before 2 August 2026 benefit from a grace period until 2 December 2026.
Sources
Would you like to explore this topic for your company?
Check Availability