All articles
AI for business11 min read

The EU AI Act in MedTech: What medical technology companies need to know and do in 2026

The EU AI Act in MedTech: What medical technology companies need to know and do in 2026

Note: This article is for general information only and does not replace legal advice. The legal situation regarding the EU AI Act is constantly evolving; this article reflects the situation as of June 2026.

In short: The EU AI Act has applied since 1 August 2024 and is being phased in step by step. For most medical technology companies, the first lever is not the much-cited high-risk obligations, but two rules that have already applied since February 2025: the ban on certain AI practices (Art. 5) and the obligation to ensure AI literacy among employees (Art. 4). The high-risk requirements have been postponed by the “Digital Omnibus” (political agreement on 7 May 2026): to December 2027 (stand-alone systems) or August 2028 (AI in regulated products such as medical devices) (Council of the European Union, 2026). If you inventory your AI applications now, classify the risk and document it verifiably, you will meet the obligations that apply today and be prepared for those still to come.

Hardly any topic is currently causing as much uncertainty in the medical technology industry as the EU’s AI Act. The headlines swing between “€35 million fine” and “everything postponed”. Both are true, and both are misleading if you only read the headline. This guide explains what the EU AI Act means in concrete terms for MedTech and medical devices, which deadlines really matter and what you should do now.

What is the EU AI Act? A simple explanation

The EU AI Act (officially Regulation (EU) 2024/1689) is the world’s first comprehensive law regulating artificial intelligence. It takes a risk-based approach: it is not AI as such that is regulated, but the specific use case. The higher the risk to people and fundamental rights, the stricter the obligations (European Commission, n.d.).

The regulation has been in force since 1 August 2024 and applies directly throughout the EU. However, its obligations do not take effect all at once, but are staggered over several years.

Which risk classes does the EU AI Act define?

The EU AI Act divides AI systems into four classes. The classification determines the entire effort involved (European Commission, n.d.):

Unacceptable risk (prohibited): Practices such as subliminal manipulation, social scoring by public authorities or emotion recognition in the workplace. These have been prohibited since February 2025.

High risk: AI in sensitive areas (e.g. critical infrastructure, recruitment, education) or as a safety component in regulated products. Here, the full catalogue of obligations applies (risk management, technical documentation, human oversight, conformity assessment).

Limited risk: Systems such as chatbots or assistants with transparency obligations under Art. 50. Users must be able to recognise that they are interacting with AI or that content is AI-generated.

Minimal risk: The majority of today’s applications, such as spam filters. No special obligations.

What matters in practice: most AI tools that medical technology companies use today, from ChatGPT to specialised knowledge platforms, fall into the limited or minimal risk categories. Genuine high-risk obligations mainly affect those who build AI into medical devices, and in medical technology that is a realistic scenario (more on this below).

Does the EU AI Act apply to my company if we only use AI?

Yes. The regulation distinguishes four roles (provider, deployer, importer and distributor), each with different obligations. As users of internal tools, most medical technology companies are deployers: they do not develop the AI themselves, but use tools. Deployers have a reduced catalogue of obligations (Art. 26), but that does not mean they have no obligations (European Commission, n.d.).

The requirement most often underestimated in practice is Art. 4 (AI literacy). Since 2 February 2025, it has obliged all providers and deployers to ensure that employees who operate AI systems have sufficient AI literacy. This expressly also applies to teams in sales, regulatory affairs, quality management or development that use tools such as ChatGPT, Copilot or dedicated specialist applications. According to specialised law firms, anyone who has not documented anything on this is already in breach of their obligations today (cf. articles by specialised law firms on the Digital Omnibus, 2026).

Which deadlines really apply in 2026 and 2027?

This is where the biggest misunderstanding lies. The original deadline of 2 August 2026 was relaxed for some of the obligations by the Digital Omnibus (political agreement between the Council and Parliament on 7 May 2026) (Council of the European Union, 2026). The current status (June 2026):

  • Prohibited practices (Art. 5) and AI literacy (Art. 4): unchanged, applicable since 2 February 2025.
  • Obligations for general-purpose AI models (GPAI): applicable since 2 August 2025, enforcement from 2 August 2026.
  • Transparency obligations for deployers (Art. 50): applicable from 2 August 2026.
  • Labelling of AI-generated content by providers (Art. 50(2)): postponed to 2 December 2026.
  • High-risk systems under Annex III (stand-alone): postponed to 2 December 2027.
  • High-risk AI in regulated products under Annex I (e.g. medical devices): postponed to 2 August 2028.

Three points are critical to understand:

  1. The Omnibus expressly does not change Art. 4 and Art. 5. These obligations have applied since February 2025 without a transition period.
  2. Only part of Art. 50 was postponed. The only thing postponed is the labelling/watermarking obligation for model providers (Art. 50(2)). The transparency obligations for deployers, and therefore for most medical technology companies, have applied as normal since 2 August 2026.
  3. The postponement of the high-risk deadlines is a breathing space, not an all-clear. Experience shows that a conformity assessment takes three to six months. If you are affected, you should use the time gained rather than wait.

And in general: the legal situation is in flux. Individual deadlines may shift further; the direction of the requirements will not change. (This article reflects the situation as of June 2026.)

What does the EU AI Act mean specifically for medical technology?

For medical technology companies, there is a special feature that is often overlooked in general guides. Medical devices and in vitro diagnostic medical devices are products regulated under Annex I. If you integrate AI into a medical device as a safety component, or use AI that is itself a medical device (software as a medical device), for example for AI-assisted reporting of imaging data or to support a diagnosis, you are therefore dealing with high-risk requirements.

What matters is which set of rules these requirements run through. Medical devices are already subject to the Medical Device Regulation (MDR, Regulation (EU) 2017/745), and in vitro diagnostic medical devices to the IVDR (Regulation (EU) 2017/746). The EU AI Act does not require you to go through a second, separate conformity procedure for AI in such products. The AI-specific high-risk requirements (risk management, technical documentation, human oversight) are integrated into the conformity assessment that is required anyway under the MDR or IVDR, and are also reviewed by the notified body (cf. articles by specialised law firms on the Digital Omnibus, 2026). Following the Digital Omnibus, the relevant deadline for the high-risk obligations remains 2 August 2028.

For medical technology companies, this means: no duplicate compliance under the AI Act and the MDR or IVDR in parallel, but an integrated path via the medical device legislation you already know. This is not an all-clear, because the requirements for safety and clinical evaluation remain demanding; however, the evidence moves into a regime you are already familiar with. In addition, the term “safety component” has been defined more narrowly: AI functions that merely support users or optimise performance do not automatically fall under the high-risk obligations, as long as their failure does not create health or safety risks.

This must be clearly distinguished from the far more common case: AI in the office and in day-to-day regulatory work, such as research in QM and regulatory documents, document analysis, review of technical documentation and onboarding. These internal applications are generally limited or minimal risk. Here, it is not about conformity assessments, but about transparency, traceability and demonstrating AI literacy.

EU AI Act and GDPR: What is the difference?

The two sets of rules are often confused, but they regulate different things. The GDPR protects personal data. The EU AI Act regulates AI systems, regardless of whether personal data is processed. Where an AI processes personal data, both regulations apply in parallel. For medical technology companies, this means: GDPR-compliant data processing is necessary, but not sufficient; the AI Act obligations come on top. This is particularly relevant for health data, because special categories of personal data under Art. 9 GDPR are involved here.

In Germany, the KI-Marktüberwachungs- und Innovationsförderungsgesetz (KI-MIG, AI Market Surveillance and Innovation Promotion Act) sets out the details of national implementation. It designates the Bundesnetzagentur (BNetzA, Federal Network Agency) as the central market surveillance authority, point of contact and complaints body, with the coordination and competence centre (KoKIVO) based there, and it governs sanctions. Sectoral authorities remain responsible in their areas, for example the BfArM (Federal Institute for Drugs and Medical Devices) for medical devices; the BSI (Federal Office for Information Security) contributes its cybersecurity expertise in a supporting role (Bundesministerium für Digitales und Staatsmodernisierung [BMDS], 2026; TÜV Rheinland Consulting, n.d.).

What happens in the event of infringements?

Fines are staggered according to the severity of the infringement: the use of prohibited practices can result in fines of up to €35 million or 7% of worldwide annual turnover. Infringements of the high-risk obligations can be punished with fines of up to €15 million or 3% of turnover. In each case, the higher amount applies (European Commission, n.d.).

Checklist: What medical technology companies should do now

Regardless of the postponed deadlines, preparing is worthwhile in any case. These five steps are a pragmatic way to start:

  1. Create an AI inventory. Which AI tools are actually used in the company, including “unofficially” in individual teams?
  2. Classify the risk. Assign each system to one of the four risk classes. Important: even if a system is not high-risk, Art. 6(3) requires a documented justification.
  3. Build and demonstrate AI literacy. Training, internal guidelines and a documented training history fulfil the Art. 4 obligation.
  4. Embed governance. Define responsibilities, approval processes and transparency rules (labelling of AI-generated content). In MedTech, ideally integrated with the existing QM system under ISO 13485.
  5. Choose compliant tools. Prefer providers that build data processing, source references and transparency in from the outset; this considerably reduces your own documentation effort.

How does MAIA support AI use that complies with the AI Act?

As an AI-powered knowledge platform for regulated industrial and medical technology companies, MAIA is designed for exactly this standard. Three points contribute directly to the requirements of the EU AI Act:

Traceability instead of a black box. Every answer is backed by a source reference down to the document and version. This fulfils the transparency principle of Art. 50 and makes results auditable. That is a clear difference from generic AI tools that “sound right” but cannot be backed up, and it fits the documentation culture in MDR and ISO 13485 environments.

GDPR-compliant, no training on customer data. Data processing to European standards, hosted in Germany and Switzerland. Your internal documents do not flow into third-party model training.

Controlled, documentable use. A clearly defined use case (access to knowledge rather than clinical decision-making or controlling a medical device) makes it easier to classify the risk as limited, and therefore makes your own documentation easier.

If you want to use AI in regulatory and everyday knowledge work without taking on new compliance risks, these are exactly the characteristics to look for.

Frequently asked questions about the EU AI Act

Does the EU AI Act already apply? Yes. The regulation has been in force since 1 August 2024. The ban on certain practices (Art. 5) and the obligation to ensure AI literacy (Art. 4) have applied since 2 February 2025. The high-risk obligations take effect later (December 2027 or August 2028).

Is my medical technology company affected by the EU AI Act if we only use ChatGPT or similar tools? Yes, as a deployer. Above all, you must ensure that your employees have AI literacy, and you must not use any prohibited practices. You only have to deal with the actual high-risk requirements if you integrate AI into a medical device as a safety component or if the AI is itself a medical device; these are reviewed through the conformity assessment under the MDR or IVDR (deadline 2 August 2028).

When do high-risk AI systems in medical technology have to be compliant? Following the Digital Omnibus, the deadline for AI in regulated products (Annex I), which include medical devices and in vitro diagnostic medical devices, is 2 August 2028; the AI-specific requirements are integrated into the conformity assessment under the MDR or IVDR. For stand-alone high-risk systems (Annex III), the deadline is 2 December 2027.

What is the difference between the EU AI Act and the GDPR? The GDPR protects personal data; the EU AI Act regulates AI systems independently of this. If an AI processes personal data, both sets of rules apply in parallel.

How high are the penalties for infringements? Up to €35 million or 7% of worldwide annual turnover for prohibited practices, up to €15 million or 3% for infringements of the high-risk obligations.

Sources

Bundesministerium für Digitales und Staatsmodernisierung. (2026, February). Kabinett beschließt schlanke KI-Aufsicht in Deutschland [Cabinet approves lean AI supervision in Germany].

European Commission. (n.d.). Regulatorischer Rahmen für künstliche Intelligenz [Regulatory framework for artificial intelligence].

Articles by specialised law firms on the Digital Omnibus on AI. (2026, May to June). [Practice articles/client briefings].

Council of the European Union. (2026, 7 May). Künstliche Intelligenz: Rat und Parlament einigen sich auf Vereinfachung der Regeln [Artificial intelligence: Council and Parliament agree on simplifying the rules] [Press release].

TÜV Rheinland Consulting. (n.d.). KI-MIG Deutschland: Behörden, Bußgelder, Fristen [KI-MIG Germany: Authorities, fines, deadlines].

Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017 on medical devices (MDR). (2017). Official Journal of the European Union.

Regulation (EU) 2017/746 of the European Parliament and of the Council of 5 April 2017 on in vitro diagnostic medical devices (IVDR). (2017). Official Journal of the European Union.

Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (AI Act). (2024). Official Journal of the European Union.

More articles

Let’s get started.

In a first intro call, we’ll work out together what a good start with MAIA looks like for you.

Cookie settings

We store your choice for 12 months.

Necessary

Security, forms and storing your choice.

Always on

Marketing

Measures whether our Google ads lead to enquiries. Google Tag Manager, Google Ads. No remarketing.

privacy policy