Artificial intelligence (AI) gets a lot of attention in health care. Some of it is hype. But in a few areas, AI tools are already approved, sold and used every day. Looking at what regulators have cleared is one of the clearest ways to see where AI is actually working.
Where AI is already in use
The US Food and Drug Administration keeps a public list of AI-enabled medical devices it has authorised [1]. A 2026 study of that list counted 1,430 authorisation records from 1995 to the end of 2025 [2]. Of these, 76.5% were in radiology. Radiology, heart and brain devices together made up 90.6% [2].
Most cleared tools help read images: flagging possible bleeding on brain scans, finding lung nodules, measuring the heart on ultrasound, or sorting urgent cases to the top of a list. Others analyse heart rhythm strips or brain signals.
What the evidence shows
An FDA clearance means a tool met the agency's standard for its intended use. It does not always mean it was tested in large trials showing better outcomes for patients. Many tools are cleared by showing they are similar to an existing device. Hospitals should look for studies done in settings like their own, with patients like theirs.
Areas where AI tools are most often used today:
- Imaging: help with reading X-rays, CT, MRI and mammograms.
- Triage: moving the most urgent scans or cases to the front of the queue.
- Heart signals: spotting irregular rhythms on ECGs and wearables.
- Paperwork: help with drafting clinical notes and coding.
Where it falls short
AI learns from data. If the data leave out some groups, the tool may work less well for them. It can also stop working as well when it is moved to a new hospital, a new scanner or a new patient group. Tools can be hard to explain, which makes it hard for doctors to trust or check them.
WHO's six principles
In 2021, WHO published guidance on the ethics and governance of AI for health. It set out six principles [3]:
- Protect human autonomy.
- Promote well-being, safety and the public interest.
- Ensure transparency, explainability and intelligibility.
- Foster responsibility and accountability.
- Ensure inclusiveness and equity.
- Promote AI that is responsive and sustainable.
In January 2024, WHO added guidance on large multi-modal models, the kind of generative AI behind chatbots. It has over 40 recommendations for governments, technology companies and health care providers [4].
What buyers should ask
- Is the tool cleared or approved for this exact use, in this country?
- Was it tested on patients like ours? How well did it do?
- How will we check that it keeps working after it goes live?
- Who is responsible if it is wrong?
- How is patient data stored and protected?
- Does it fit into the systems staff already use, or add extra steps?
Key numbers at a glance
- 1,430 FDA AI device authorisation records, 1995–2025.
- 76.5% in radiology.
- Radiology, heart and brain together: 90.6%.
- WHO's AI ethics guidance: 6 principles (2021).
- WHO's generative AI guidance: over 40 recommendations (2024).
Go deeper
Want the full picture? Our report covers market size, growth, segments, countries and the key companies. Read the report or get a free sample.
Generative AI in health care
Large language models and other generative AI tools can write text, answer questions and summarise records. WHO calls the wider group "large multi-modal models", because they can take in text, images and other data [4]. WHO's 2024 guidance lists possible uses: helping with diagnosis, guiding patients, doing office tasks, teaching health workers, and supporting research [4]. It also warns of risks: false or biased answers, privacy problems and over-reliance by users.
WHO's guidance builds on its six principles and asks governments to check these systems before wide use, and to involve doctors, patients and the public in their design [4].
Data and privacy
AI needs large amounts of health data. That raises questions about who owns the data, how patients give consent, and how it is kept safe. Clear rules on data use, strong security and the option for patients to opt out help build trust. WHO's principle of protecting human autonomy includes privacy and informed consent [3].
Who benefits
AI could widen access where specialists are scarce. But tools trained mostly on data from rich countries may work less well elsewhere. WHO's principle of inclusiveness and equity asks that AI be designed for the widest possible use, regardless of age, sex, income or other traits [3].
Signs a tool is ready for wide use
- Clear approval for its intended use.
- Published results from real hospitals, not only test data.
- A plan to watch performance over time.
- Clinicians who understand what it can and cannot do.
- A simple way to report errors.
Where to read more
All figures in this article come from public sources listed below. For more detail, start with the US FDA list of AI-enabled medical devices and WHO's 2021 and 2024 guidance on AI ethics and governance. Each is free to read online.
What this means
AI is working best today in narrow, well-defined tasks, especially reading images and sorting work. Wider uses, like diagnosis across many diseases or running whole care pathways, are earlier. The winners will be tools that prove their value with real patients, fit into daily work, and earn trust through clear rules.
Explore on Acuitas
- Artificial Intelligence (AI) in Healthcare market report – size, growth and key players
- Artificial Intelligence (AI) in Healthcare statistics – key numbers, trials and research in one place
- Top 10 fastest-growing Digital Health markets
- Global Artificial Intelligence (AI) in Drugs Discovery Market Analysis – related market
- Global Artificial Intelligence (AI) in Life Sciences Market Analysis – related market
- Global Artificial Intelligence (AI) in Medical Imaging Market Analysis – related market
- Get a free sample report or talk to our team
Topics: Artificial intelligence · Digital health · Health IT and data
Sources
- US FDA: Artificial intelligence-enabled medical devices (public list)
- Golshani and Joseph (Cureus, 2026), via BioSpace: three in four FDA AI device authorizations were in radiology
- WHO (2021): Ethics and governance of artificial intelligence for health
- WHO news (18 Jan 2024): AI ethics and governance guidance for large multi-modal models
Banner image: Female doctor in blue uniform working on laptop closeup by Shixart1985, CC BY 2.0, via Wikimedia Commons.
The data behind this story
Topics: Artificial intelligence · Digital health · Health IT and data
Go deeper