Research · Library
Analysis of AI regulation, written for the people who have to act on it.
Latest
LibraryStandard Intelligence
Which high-risk AI can be self-assessed, and which needs a notified body? This paper maps the pathways category by category, with the declaration of conformity and post-market duties.
LibraryStandard Intelligence
Article 72 requires a post-market monitoring plan for high-risk AI. Done well, it reuses the infrastructure you already run rather than spawning a second compliance programme.
LibraryStandard Intelligence
Annex IV is not a one-off deliverable. It is a living technical file across fourteen elements, and the standard notified bodies expect is higher than the text suggests.
LibraryStandard Intelligence
Article 14 requires effective human oversight of high-risk AI. The hard part is designing oversight that is genuine rather than nominal, without making the system useless.
LibraryStandard Intelligence
Article 4 has been enforceable since February 2025. Sufficient AI literacy is contextual, not universal, and a tiered, role-based programme is the only model that holds up.
LibraryStandard Intelligence
Article 5's eight prohibitions have been in force since February 2025, as absolute bans. This paper maps where the boundaries sit, including the hardest provisions to read.
LibraryStandard Intelligence
Classification is the first legal obligation under the Act, and every later decision flows from it. This decision-maker's guide maps the four tiers and the judgement calls that decide them.
LibraryStandard Intelligence
Classification is the first legal obligation and the costliest to get wrong. This guide maps the four tiers, the Article 6(3) exception, and the modern edge cases that break old frameworks.
LibraryStandard Intelligence
A whole-regime view of the EU AI Act for boards and compliance leaders: risk classification, the eight high-risk pillars, GPAI duties, enforcement, and a five-action roadmap.
Your next step