Practical guidance on EU AI Act compliance for businesses with customer-facing AI systems.
If your AI helped decide something that significantly affects a person, Article 86 of the EU AI Act lets them ask you, the deployer, for a clear explanation. How the right to explanation works, who it falls on, and how it differs from GDPR.
Read more →If you build a high-risk AI system outside the EU and sell it in, Article 22 of the EU AI Act makes you appoint an EU authorised representative before the system goes on the market. They are more than a mailbox — they can end the mandate and report you.
Read more →Article 15 makes you commit accuracy thresholds to writing, then expects the system to keep hitting them. It also names the cybersecurity attack classes your system has to resist.
Read more →Article 13 of the EU AI Act requires high-risk AI systems to be designed so deployers can interpret their output, not just documented after the fact. What transparency by design means and what the instructions for use must contain.
Read more →After the April trilogue collapsed, negotiators returned and struck a deal. The Annex III high-risk deadline officially moves from 2 August 2026 to 2 December 2027.
Read more →Article 11 turns your technical documentation into a regulated deliverable. It has to be drawn up before launch, kept current as the system evolves, and ready to hand to an inspector on demand.
Read more →Article 10 of the EU AI Act governs the data behind high-risk AI: design choices, sourcing, bias examination, and data gaps. What the provider must document, and how the record survives an audit.
Read more →Superseded — the May 2026 trilogue later reached a deal moving the Annex III high-risk deadline to 2 December 2027. This is the 6 May analysis from when the April talks had collapsed and 2 August 2026 still stood.
Read more →Most teams think Article 50 is an afternoon's work — add an 'AI' badge to the chatbot, done. That covers one of four obligations. The machine-readable marking of synthetic content, the 'obvious AI' exception, and what the Commission's draft guidelines now expect. Disclosure applies from 2 August 2026.
Read more →Article 14 of the EU AI Act requires that humans can effectively oversee high-risk AI systems — a design bar, not a staffing one. Why most 'human in the loop' setups fail the effectiveness test, and the four things real oversight needs.
Read more →