European Union: Implemented AI Act including design requirement for Annex III "high-risk AI systems"

On 2 August 2026, the regulation laying down harmonised rules on Artificial Intelligence (Artificial Intelligence Act) comes into effect for "high-risk AI systems" referred to in Annex III of the Act. The Act includes design requirements for "high-risk AI systems" as part of a wider requirement for a risk management system (Art. 9). Specifically, identified risks should, as far as possible, through the implementation of technical measures during the design and technical process, be either eliminated or mitigated and controlled. Further, high-risk AI systems should be designed such that they can be used transparently, with a sufficient level of human oversight, and with the capacity to automatically record their use. The Act requires AI providers to generate synthetic content that must ensure outputs are machine-readable and marked as artificial. The marking must be effective, interoperable, and technically feasible, considering content type, costs, and technical standards. The Act includes exemptions for minor edits, assistive functions, or lawful criminal investigations. The Act is generally implemented on 2 August 2026 after the lapse of the grace period, though some sections are subject to variable implementation periods. For high-risk AI systems as defined by Article 6(1), the Act only applies from 2 August 2027.