Using AI for order emails, quotes, planning suggestions and reports is allowed, provided a person checks the result. That check is the same thought as in Article 4. At a wholesaler, nobody has to retype two out of three order emails any more (3,300 of 5,017 order emails, measured in September 2026): before, everything was by hand, now 67.7% arrives as a correct draft order. Exceptions go to an employee.
Transparency rules have applied since 2 August 2026: anyone who shows customers a chatbot or AI-made images should make clear that it is AI.
High-risk AI means AI that influences decisions about people or safety, such as selecting and assessing staff or judging creditworthiness. Those rules were postponed by the omnibus: to 2 December 2027 for uses of this kind, and to 2 August 2028 for AI in products already covered by other product law. Whoever builds an AI system for others has more duties than whoever only uses one; a wholesaler or manufacturer usually falls in the second group.