The way the European market handles AI health tools is just different. A notified body is your gatekeeper to market access, full stop. This framework, laid out in the MDR EU Regulation, means manufacturers must have an intense documentation discipline. For any quality and regulatory team, this means getting ready for Europe isn’t about having a slick product. It’s about proving the completeness and accuracy of your technical file.
Companies like Viz.ai, Butterfly Network, and Tempus AI are all working through both the US FDA and European regulatory fields, so they have to build a single conformity thread that satisfies both systems. While the FDA cleared 295 AI/ML devices in 2025, with 96% taking the 510(k) pathway FDA AI/ML medical device clearances, the European approach to conformity assessment piles on a layer of scrutiny that feels much heavier, especially for high-risk AI applications in healthcare.
The Notified Body as Gatekeeper
In the EU, a notified body is an independent organization that a member state picks to assess products before they hit the market. For medical devices, AI-powered ones included, these bodies are the ones who ensure you’re compliant with the MDR EU Regulation. You’ll see names like BSI Group and TUV SUED pop up as the primary notified bodies for the AI Act and MDR. Their involvement is a substantive review of your technical documentation and quality management system, not a rubber stamp.
What the notified body does is different from the FDA’s premarket review, even if they both want safe and effective devices. The European system is obsessed with a manufacturer’s ability to show ongoing conformity through the entire product lifecycle. This means you need a strong Quality Management System (QMS) that’s compliant with standards like ISO 13485, which, let’s be honest, the FDA is expecting more and more as well ISO 13485 QMS standard.
The Technical File: A Structured Argument for Conformity
Your technical file is the complete dossier you hand over to the notified body for review, and it has to systematically address every relevant part of your device, from design and manufacturing to clinical performance and post-market plans. The real insight here is that conformity review is a document discipline. The assessor literally works through your technical file, and you can judge your own readiness by asking if that file cleanly answers questions about classification, evidence, and post-market duties, in that specific order.
Device Classification and Intended Purpose
The first thing to get right in the technical file is the classification of your AI health tool. Under the MDR EU Regulation, AI devices often get a high-risk classification, which immediately triggers tougher conformity assessment procedures and dictates exactly which modules and requirements the notified body is going to pore over. You have to define the intended purpose of the device with absolute clarity, because that definition is the foundation for your entire regulatory strategy and determines the scope of clinical evidence you’ll have to produce.
Evidence of Safety and Performance
You must provide complete clinical evidence showing your AI health tool is safe and performs as advertised. This isn’t just about your big clinical trial. It includes data from scientific literature and, increasingly, real-world evidence (RWE). This RWE, which can come from electronic health records, registries, or claims data, can be used to supplement your main trials and strengthens both your FDA submission and your story for payers Real-world evidence for medical devices. The quality of this evidence is everything. The notified body will pick apart your clinical data to see if it actually supports your claims and if you’ve accounted for AI-specific risks like algorithmic drift.
Post-Market Surveillance and Change Control
The technical file has to detail your post-market surveillance (PMS) plan. This means spelling out procedures for vigilance, post-market clinical follow-up (PMCF), and continuous risk management. With AI health tools, this is especially important because the models themselves are dynamic. The EU AI Act, with its new compliance deadlines for high-risk AI now mostly in December 2027 or August 2028, just puts more pressure on having strong monitoring and governance in place EU AI Act compliance deadline.
Your file also has to outline your change control processes. AI/ML models can adapt and learn in ways that traditional medical devices can’t. A Predetermined Change Control Plan (PCCP), an FDA concept, but the principle is universal, is a way to manage changes to your AI device without having to do a whole new premarket submission for every single update. Without a plan like this, you’d be looking at a new submission every time your model retrains on new data, which is completely unscalable.
Conformity in Practice: Viz.ai, Butterfly Network, and Tempus AI
The public conformity materials from companies like Viz.ai, Butterfly Network, and Tempus AI show how this works on the ground. As vendors who have to operate in both the US and Europe, they understand that regulatory compliance is an ongoing commitment to documentation and quality. Their materials all follow the same conformity thread, showing they get the dual compliance requirements for any AI health company trying to go global Dual compliance requirements for AI health companies.
Take Viz.ai, a company that comes up constantly in these discussions. They have to make sure their technical documentation works for both the FDA’s SaMD principles and the MDR EU Regulation. It’s the same story for Butterfly Network and Tempus AI, whose products rely heavily on AI. The fact that you see these same companies mentioned in articles about FDA risk scores and in EU compliance guides shows just how interconnected global AI health regulation has become.
Judging Readiness Before Assessment
For any quality and regulatory team, being able to accurately judge your own readiness before you even talk to a notified body is a huge advantage. It’s about doing a cold, hard review of your technical file against the MDR EU Regulation’s requirements. Can you honestly say the file answers the mail on classification, evidence, and post-market duties, in that order? If you can, you’re probably in good shape.
This internal audit means confirming your device classification is defensible, your clinical evidence is solid and directly backs up your intended purpose, and your post-market plan is actually doable and accounts for the weirdness of AI. It also means your QMS is fully up and running and you can prove it’s effective. A clean data room, with organized FDA correspondence, your SOC 2 reports, and key customer contracts all lined up, is the sign of a mature company that’s ready for someone to come look under the hood.
The notified body review is just an independent check on the claims of conformity you’ve already made. It’s a process that rewards people who are thorough and proactive. For AI health tools, where the tech moves so fast, this structured assessment is a necessary check to make sure those advances result in safe and effective medical devices for patients in Europe.
Frequently Asked Questions
What is the role of a Notified Body in the European regulatory process for AI health tools?
In the European Union, a Notified Body is an independent organization designated by an EU member state to assess the conformity of products, including AI-powered health tools, with the MDR EU Regulation. They serve as gatekeepers for market access by conducting a substantive review of a manufacturer’s technical documentation and quality management system.
What is the ‘technical file’ and why is it important for European submissions?
The technical file is the core document manufacturers submit to a Notified Body for review. It is a comprehensive dossier that systematically addresses all relevant aspects of the device, from design and manufacturing to clinical performance and post-market surveillance. Its completeness and accuracy are crucial for demonstrating conformity with the MDR EU Regulation.
What key elements must be included in the technical file for an AI health tool?
The technical file must include the accurate classification of the AI health tool and its clearly defined intended purpose. It also requires comprehensive clinical evidence demonstrating safety and performance, and a detailed post-market surveillance plan, including procedures for vigilance, post-market clinical follow-up, and continuous risk management.
How does the European regulatory approach differ from the US FDA’s for AI health tools?
While both aim for safety and efficacy, the European system, governed by the MDR EU Regulation, places a heavy emphasis on a manufacturer’s ability to demonstrate ongoing conformity throughout the product lifecycle, not just at market entry. It often adds a layer of scrutiny perceived as more demanding, particularly for high-risk AI applications, and requires involvement of a Notified Body.