Voluntary test seals for AI systems are currently being developed, tiered by depth of assessment and available from late 2026. The need is real, as questions about AI usage appear ever more frequently in tenders and supplier questionnaires.
One point is decisive: such a seal is voluntary and does not replace the statutory conformity assessment, it adds technical assessment aspects. For Swiss companies this matters, because pressure here rarely originates from a regulator. It comes from customers in the European market who require evidence.
In customer conversations the question is therefore rarely about a seal, but whether you can explain what a system does, which data it works with, and who is accountable for a result. The preparatory work is worthwhile regardless of any seal: an inventory of the systems in use including process and owner, a statement on the origin of the models, a written boundary for human decisions, and logging that remains meaningful months later.