Life Sciences & Health Care 03.04.2026

5 Min. read time

By the end of 2025, the US FDA had approved exactly 1,451 AI-powered medical devices – 295 of them in the past year alone. In Europe, there isn’t even a comparable public database. While US clinics are using AI diagnostics as part of their standard operations, German hospitals are working with isolated solutions and waiting for an AI Act, whose high-risk requirements won’t come into effect until 2027. A stocktake.

What are AI medical devices?

AI medical devices are a concrete lever for companies in 2026 because the topic directly affects digital care processes, data protection, and hospital IT. The article uses the example of synaforce to show which requirements, metrics, and operational steps count in practice.

295 approvals in one year: What the FDA figures show

Since 2017, the FDA has maintained a publicly accessible database of all AI-powered medical devices with market approval. The figures show a clear trend: from 29 approvals in 2017 to 295 in 2025 – a record figure that significantly exceeded the previous year’s record of 223.

76% of all approved AI devices are for radiology – an area where pattern recognition works particularly well with large image datasets. GE HealthCare tops the list with 120 approvals, followed by Siemens Healthineers with 89. The fact that two of the top three manufacturers have European roots makes the regulatory asymmetry between the US and EU even more striking.

However, not every approval is equivalent to clinical benefit. An analysis in JAMA Network Open (April 2025) examined 903 FDA-approved AI medical devices and came to a sobering conclusion: 24.1% had no clinical validation study to speak of. Only 2.4% were tested with randomized controls, 8.1% prospectively. Around 90% of the devices relied on retrospective data or no clinical evidence at all.

This means that the FDA figure of 1,451 approvals is impressive as an indicator of market dynamics. As a quality characteristic, it is only conditionally suitable. For hospital buyers evaluating an AI system, FDA approval alone is not a sufficient criterion. What matters are prospective studies on comparable patient cohorts and comprehensible clinical validation in their own care context.

Europe’s Blind Spot: No Database, No Transparency

In Europe, there is no comparable public list. Neither the EMA nor the BfArM maintains an accessible database of certified AI medical devices. The European database EUDAMED, intended to provide transparency, has been under development for years. Individual modules are set to become mandatory from May 2026, but a comprehensive overview of all AI products on the European market will not be available until years later.

Independent trackers estimate the number of CE-certified AI medical devices to be at least 219 (as of January 2025). The actual number is likely higher, but no one knows for sure. For hospital procurement managers deciding which AI system to evaluate, this is a real problem.

DACH Clinics Deploy AI – But in Isolated Solutions

Despite regulatory gaps at the European level, individual DACH clinics are already productively using AI diagnostics. The results are promising, but remain isolated.

Asklepios, one of Germany’s largest private hospital operators, has installed the Israeli AI system Aidoc in over 25 facilities. The software analyzes CT scans in real-time and detects strokes, intracranial hemorrhages, and vertebral fractures – funded through KHZG subsidies. The advantage: Aidoc prioritizes the findings list, so time-critical cases immediately land with the radiologist instead of in the queue.

The largest prospective study on AI use in diagnostics to date also comes from Germany. The Vara-PRAIM study, published in Nature Medicine, evaluated over 463,000 mammography screenings. The result: a 17.6% increase in breast cancer detection while simultaneously reducing the workload for radiologists. It is the world’s largest prospective study of its kind.

“The question is no longer whether AI works in diagnostics. The question is how we achieve the transfer from the study to routine operation – regulatory, organizational, and economically.” – Prof. Alexander Berens, University of Tübingen (press release, 2025)

The problem: Each of these implementations is a standalone project. There is no national standard for AI integration into hospital information systems, no uniform interface, and no central registry of which clinic uses which system. Although the G-BA Innovation Fund is promoting the regular use of AI in radiology with the xR.AI project, Germany is far from a comprehensive rollout.

The consequence: German clinics that want to introduce AI diagnostics must navigate every step themselves – from product selection to integration into their HIS to training radiologists. In addition, there are liability questions that have not yet been conclusively clarified nationally or at the European level. If an AI system misses a finding and the treating physician has relied on the system, who bears the responsibility? The MDR regulates the manufacturer’s product liability, but the medical duty of care remains. In practice, this uncertainty means that many radiologists do not use AI findings as a basis for decision-making, but at best as an additional hint.

AI Act from 2027: What’s in Store for MedTech Companies

The EU AI Act has been in force since August 2024, but medical devices have a transition period: the full high-risk requirements will only apply from August 2, 2027. AI systems classified as medical devices under the MDR or IVDR are automatically categorized as high-risk AI.

What this means in concrete terms: manufacturers must implement an AI-specific risk management system, meet data quality and data governance requirements, create technical documentation that goes beyond MDR requirements, and integrate human oversight systems. The effort is significant, particularly for smaller companies.

A glimmer of hope: in November 2025, the EU Commission presented an omnibus simplification proposal aimed at reducing the overlap between the AI Act and MDR/IVDR. Instead of duplicate documentation and conformity assessments, there will be an integrated process. Adoption is expected in summer 2026.

For German MedTech SMEs – 93% of the industry are companies with fewer than 250 employees – this is not an academic question. The MDR has already led to manufacturers withdrawing products from the European market in recent years because certification costs exceeded revenue. If the AI Act adds similar bureaucracy, consolidation will continue.

At the same time, regulation creates a quality advantage: AI medical devices that comply with both MDR and AI Act have a trust advantage over devices approved solely by the FDA. European manufacturers that overcome this double hurdle can use it as a differentiator in the global market. Prerequisite: they must survive the transition phase until 2027 financially and organizationally.

Gaining Momentum or Losing Ground

The global AI market in healthcare was estimated at $21.66 billion in 2025. By 2030, it’s expected to grow to over $110 billion – an average annual growth rate of 38.6% (MarketsandMarkets). PwC Strategy& estimates the European potential at $203 billion.

Germany is home to strong AI players in MedTech, including Siemens Healthineers, Brainomix, Vara, and Merantix. Clinical research is on an international level – the PRAIM study is a striking example. What’s missing is the bridge between research and routine operation: a European AI database that creates transparency; a harmonized approval process that doesn’t overwhelm SMEs; and a hospital landscape that sees AI not as a funding project, but as a standard tool.

The FDA has approved 1,451 AI medical devices in eight years and has learned to adapt its processes. Europe has written a regulation in the same timeframe, which will only take effect in 2027. Anyone who doesn’t see this backlog as a strategic risk hasn’t understood the market’s dynamics.

Frequently Asked Questions

How many AI medical devices has the FDA approved by 2025?

By the end of 2025, the FDA had approved a total of 1.451 AI-powered medical devices. In 2025 alone, 295 were added – a record number. 76% are in radiology, with GE HealthCare leading the list with 120 approvals.

Is there a database for AI medical devices in Europe?

No. Unlike the FDA, neither the EMA nor the BfArM maintains a public list of CE-certified AI medical devices. The European database EUDAMED is under development, with individual modules becoming mandatory from May 2026. A complete overview will not be available for years.

When do the AI Act requirements for medical devices take effect?

The full high-risk requirements of the EU AI Act for medical AI systems will apply from August 2, 2027. AI systems classified as medical devices under MDR or IVDR are automatically considered high-risk and must meet specific AI-related risk management and data quality requirements.

Which German hospitals are already using AI diagnostics?

Asklepios is using the AI system Aidoc in over 25 hospitals for CT analysis (funded by KHZG). The Vara-PRAIM study at German breast cancer centers increased the detection rate by 17.6% among 463,094 women examined. The G-BA project xR.AI promotes the regular use of AI in radiology.

How big is the market for AI in healthcare?

The global market was worth 21.66 billion US dollars in 2025 and is expected to grow to over 110 billion US dollars by 2030 (CAGR 38.6%, MarketsandMarkets). PwC Strategy& estimates the European potential at 203 billion US dollars by 2030.

Source title image: Pexels / MART PRODUCTION (px:7089623)