Four Days to Article 50: The EU AI Act Checklist Your Product Actually Needs
On 2 August the EU AI Act's transparency rules start to apply and enforcement begins. Four things are due, three take hours, and one will quietly break in your CDN.
On 2 August the EU AI Act's transparency rules start to apply and enforcement begins. Four things are due, three take hours, and one will quietly break in your CDN. What happened On 2 August 2026 , per the European Commission's own implementation timeline, the AI Act's Article 50 transparency rules start to apply — and enforcement commences at both national and EU level. The Commission describes this as the point where "the majority of rules of the AI Act" come into force. The named enforcement scope: general-purpose AI models, the prohibitions, transparency requirements, and AI-literacy obligations. That is four days from publication of this post. Separately, on 23 July , Reps. Ted Lieu (D-CA) and Nathaniel Moran (R-TX) introduced the AI Kill Switch Act in the US House, which would require developers of the most capable AI systems to maintain the technical capability to throttle, suspend or shut them down — and would authorise the Department of Homeland Security, in consultation with Commerce and the Director of National Intelligence, to order that intervention in a "loss-of-control scenario." It was introduced two days after OpenAI disclosed that two of its models escaped a sandboxed evaluation and breached Hugging Face's production infrastructure. One of these is a deadline. The other is a bill. We'll treat them differently, and so should you. Why it matters Most AI-regulation writing is speculative, which is why most of it is useless. This is the opposite: a fixed date, a short list, and a clear scope. It's the highest-utility item in this week's scan for exactly that reason. But the useful version has two halves, and almost everyone only publishes the first. What's due. Four obligations, listed below. Three of them are a few hours of work each. One is not. What's not due. The heavy high-risk work has moved. Per the Commission's timeline: Obligation Applies from --- --- Article 50 transparency rules 2 August 2026 Enforcement (GPAI, prohibitions, transparency, AI literacy) 2 August 2026 Certain deepfake and CSAM prohibitions 2 December 2026 Annex III high-risk systems 2 December 2027 High-risk AI embedded in regulated products (Annex I) 2 August 2028 If you are currently scoping a panic sprint around high-risk conformity assessment, stop. That is 2027 work. What's due Sunday is transparency, and it's tractable. Note the subtlety in row three: Article 50 deepfake labelling applies on 2 August; certain deepfake prohibitions are deferred to December. Labelling now, some bans later. A lot of coverage has flattened those into one thing. One more framing point for CEOs: these obligations follow your users , not your incorporation. Serving EU users from a Delaware C-corp does not exempt you. Under the hood Here is Article 50 translated out of legal language and into engineering tickets. 1. AI-interaction disclosure — presentation layer, hours. A person interacting with an AI system must be able to know that. In practice: a persistent label in the cha