EU AI Act delay leaves SME transparency deadline intact

The EU’s Digital Omnibus on AI has delayed some of the EU AI Act’s heavier high-risk system obligations, but it has not moved the core transparency deadline affecting many SMEs. Standalone high-risk systems under Annex III, including recruitment tools, credit scoring and education access systems, now have until 2 December 2027, while AI embedded in regulated products such as medical devices, machinery and toys has until 2 August 2028.

Article 50 transparency rules still apply from 2 August 2026. Businesses using customer-facing chatbots, voice assistants or AI agents must disclose when users are interacting with AI unless it is obvious from context. AI-generated or altered images, audio or video that could appear real must be labelled, and users exposed to biometric categorisation or emotion recognition tools must be told those systems are in use.

Watermarking rules depend on launch timing: tools launched before 2 August 2026 have until 2 December 2026 for machine-readable marking, while tools launched on or after 2 August 2026 must comply from day one. Article 50 breaches can carry fines of up to €15 million or 3% of global turnover, while prohibited practices can reach €35 million or 7%.