Is an inspection robot a high-risk AI system?

In most cases not – what matters is what the AI in the device is responsible for.

If the system uses machine learning for navigation, mapping or image analysis while the safety functions work deterministically, there is as a rule no high-risk system.

It is different if a safety function itself is based on machine learning – for example, if a learning model decides whether the device stops. A conformity assessment by a notified body is then required, and classification as high-risk becomes a real possibility.

A specific question for the manufacturer: which safety functions are based on machine learning, and which are not? More in article on the AI Act and the Omnibus amendment.