Geoffrey Ceunen’s Post

The European Commission has just published a comprehensive new resource — the AI Act Service Desk FAQ. This document brings together insights from over 100 real questions raised by organizations, lawyers, and policymakers during the AI Pact webinars — now answered directly by the Commission. Why is this such an important release? Because it’s not just another summary — it’s an official, evolving guide that helps interpret how the EU AI Act will actually work in practice. Here’s what you’ll find inside: 🔹 Clarity on which systems are in scope (and which aren’t) 🔹 A full rollout timeline through 2027 🔹 Clear definitions of AI systems, AI models, and general-purpose AI 🔹 Details on obligations for high-risk AI and prohibited practices 🔹 Guidance on the General-Purpose AI Code of Practice and regulatory sandboxes 🔹 Even technical criteria — like the FLOP thresholds (10²³ and 10²⁵) used to flag models that pose systemic risk Published by the European Commission’s AI Act Service Desk, this FAQ is now one of the most credible and practical interpretative sources for understanding the law. If your work touches AI governance, compliance, or risk — this one’s worth bookmarking. #EUAIAct #AICompliance #AIGovernance #ResponsibleAI #TechPolicy

🇪🇺 𝗡𝗮𝘃𝗶𝗴𝗮𝘁𝗶𝗻𝗴 𝘁𝗵𝗲 𝗘𝗨 𝗔𝗜 𝗔𝗰𝘁 𝗷𝘂𝘀𝘁 𝗴𝗼𝘁 𝗮 𝘄𝗵𝗼𝗹𝗲 𝗹𝗼𝘁 𝗲𝗮𝘀𝗶𝗲𝗿. The European Commission has launched the EU AI Act Service Desk — a new one-stop online platform designed to help organisations understand and comply with the world’s first comprehensive AI law. 𝗪𝗵𝘆 𝗶𝘁 𝗺𝗮𝘁𝘁𝗲𝗿𝘀 👇 🔹 The AI Act is moving from legislation to implementation — fast. 🔹 The Service Desk offers official guidance, FAQs, and direct support from the Commission. 🔹 It’s now the central hub for anyone developing, deploying, or advising on AI in Europe. If your organisation uses or builds AI systems, this is the time to start your compliance journey. 💡 Read our new blog post to learn what the Service Desk offers, who should use it, and why it marks a turning point for AI governance in Europe: 👉 https://lnkd.in/d_jYRj4E

🔍 Reflections from behind the scenes of ActiveFence’s new whitepaper on GenAI regulation I recently had the privilege of contributing to ActiveFence’s whitepaper on GenAI regulation - a deep dive into the shifting global legal terrain. What hit me hardest? The breakneck speed of change. We’d map out recent laws, frameworks, and standards -only to discover that while drafting, new obligations had landed. 🧭 So instead of asking if we should regulate AI, the question becomes: how do we keep pace -and stay ahead? In your view, what’s the single biggest legal risk or blind spot companies face when deploying AI today? 👇 The report is linked in the first comment. Dive in!

AI Compliance-by-Design: A CTO Guide to Model Inventory, RAG Sources, and Decision Logging under the EU AI Act The EU AI Act is now in force and rolling out in phases; several obligations arrive before full application in 2026–2027 (see the EU’s implementation timeline and this key-dates overview). The European Commission’s AI Office will oversee providers — with a special focus on general-purpose AI (GPAI) — and may request information or evaluate models. For GPAI in particular, the Commission published obligations fact pages and a voluntary GPAI Code of Practice to reduce ambiguity while Article 53/55 duties phase in. Below is the working checklist we use at Pynest to make AI systems shippable across jurisdictions without turning every release into a legal fire drill. The Act differentiates between providers (who place on the market or put into service) and deployers (who use AI systems). If you provide or fine-tune GPAI models, you face specific duties such as technical documentation, a copyright policy, and a summary of training content (see the Commission’s GPAI obligations; also note the https://lnkd.in/gV9YZdUb

EU AI Act: Timelines, Standards & Regulatory Readiness At a recent conference, EU digital chief Henna Virkkunen reaffirmed her commitment to keeping the AI Act on schedule, warning that delays in standard-setting risk leaving businesses without clarity on compliance. Key takeaways for the legal and AI community: Timeline pressure: Despite discussions of a potential “stop the clock” mechanism, Virkkunen insists delays could erode legal certainty. Standards lagging: European standardization bodies are behind in finalising the compliance frameworks—crucial for high-risk AI systems. National regulators: Member states are still late in setting up their competent authorities, adding to the uncertainty. Business impact: The Act’s high-risk AI requirements are set to apply from August 2026, with SMEs highlighted as a priority for support. For UK-based AI legal experts and lawyers, this is a pivotal moment: understanding EU regulatory timelines is vital for anticipating how UK frameworks may align, diverge, or respond. 📞 If you need to stay ahead in tracking UK AI regulation and its interaction with EU law, get in touch: Phone: 07552 882 850 Email: nicholas.mitchell@lexisnexis.co.uk

⚖️ Three States, Three Paths: Courts and AI Yesterday’s post zoomed in on Qld’s pragmatic approach to AI in litigation. Nationally, though, the picture is far from uniform. Within months of each other, courts in NSW, Victoria and Qld have all released their own guidance on generative AI: - NSW: 🚫 strict bans & mandatory disclosures - Victoria: 🧭 principles-based trust - Queensland: 📝 pragmatic policy The result? A patchwork of obligations. Lawyers now need to adapt state by state — and harmonisation looks a long way off. 👉 Full breakdown here: https://lnkd.in/gce2amYw #AI #DigitalLaw #Regulation #LegalTech #Litigation #Australia #IPMojo

If you work on AI Act compliance (or just need to stay in the loop) and haven’t explored the AI Act Service Desk resources yet - you really should. Beyond trying out tools like the Compliance Checker and AI Act Explorer, it’s definitely worth visiting one of the less-highlighted sections: the FAQ. Some of the questions and answers are quite basic, but others provide genuinely useful insights into topics that aren’t directly regulated under the AI Act. For example: ❓ How do AI agents fit within the General Purpose AI framework under the AI Act? ❓ Is an "AI Officer" or governance board recommended within companies? That said, even the answers to the more basic questions about the AI Act may be quite helpful - for two reasons: ▶️ First, they point to different resources on the AI Act already created by the European Commission, making the FAQ a great interactive hub for finding official information. ▶️ Second, the explanations break down AI Act concepts and requirements in a clear, concise way. It’s an excellent go-to resource whenever you need to briefly and simply explain issues such as which types of AI systems are regulated or what obligations correspond to different risk levels. The list of questions isn’t long yet, but it’s expected to be updated regularly - something to look forward to. #AIAct #AIActServiceDesk

I read this article with interest by the team Barnes & Thornburg LLP.   In short, it talks about a proposed US Federal Rule of Evidence 707 which would subject “machine-generated evidence” to the same admissibility standard as expert testimony   The piece says, “to be admissible, the proponent of the evidence must show that the AI output is based on sufficient facts or data, produced through reliable principles and methods, and demonstrates a reliable application of the principles and methods to the facts.”   I think it’s inevitable that AI-generated evidence will make it to the courtroom sooner or later (and probably has already – regardless of whether that has been disclosed or not) – probably initially as part of expert evidence, where the expert has used #AI to assist with their analysis.   This article discusses the proposed evidential rule in the US to put guardrails around the use of AI-generated evidence as a standalone piece of #legal #evidence. It’s certainly important guardrails are established so that the evidence can be scrutinised and tested to ensure it is valid in both approach, inputs and results, otherwise it could be misleading. As such, it is right that rules are in place, it is right that the AI-generated evidence can, when complying with those rules, be admitted – but the process by which it is generated and used should be fully disclosed and be open to scrutiny by the opposing side and their experts. This will lead to “battles” as the article points out, but the use of AI-generated evidence must be transparent – always. I’ll be watching with interest..

Artificial intelligence tools are already transforming the way construction and infrastructure disputes are run, from drafting submissions to reviewing thousands of documents in seconds. The Chartered Institute of Arbitrators (Ciarb) has released its 𝐆𝐮𝐢𝐝𝐞𝐥𝐢𝐧𝐞 𝐨𝐧 𝐭𝐡𝐞 𝐔𝐬𝐞 𝐨𝐟 𝐀𝐈 𝐢𝐧 𝐀𝐫𝐛𝐢𝐭𝐫𝐚𝐭𝐢𝐨𝐧 (2025), a timely and much-needed framework for practitioners, arbitrators, and parties grappling with the integration of AI into arbitral proceedings. Glaholt Bowles LLP Partner Michael Valo discussed why this matters, the key takeaways, and what the CIArb Guideline on the Use of AI in Arbitration means in practice in our Summer 2025 newsletter. Read more here: https://lnkd.in/e8zxj2FP