AI in Healthcare Regulation: Will You Be Told AI Was Used?

On 6 October 2026 the UK government accepted all 44 recommendations of its National Commission into the Regulation of AI in Healthcare.¹ The announcement included "new commitments to keep patients informed about when AI is used in their care."

It sounds like a new right. The commission's own report says that today there is "no universal requirement or expectation" for UK health professionals to tell patients each time AI is used.²

Some places already require it, for some uses. Others have nothing written down. For a patient travelling for treatment, the rule that counts is usually the one where the clinic is, not the one at home. This article sets out what each of seven jurisdictions requires as of October 2026, and what a patient can ask for where nothing is required.

The UK: a promise, and a warning against consent forms

The UK commission's report, published in September 2026, is frank about where things stand.² A majority of participants in deliberative research by the Health Foundation thought patients should be told when AI has been used in their care. Yet the report finds no universal requirement or expectation to tell them.

Its answer is Recommendation 35. Health organisations should adopt "a proportionate, system-level approach to transparency." This should address "patients' reasonable expectation of being informed" and give them "the ability to opt out where possible or appropriate."

The report rules out consent for every use. It points out that clinics do not seek explicit consent for every supporting technology, such as electronic patient records. Requiring it for every AI use "would be impractical," the commission writes, and could harm the delivery of care. It proposes general public notice, with more detailed discussion kept for high-risk or novel uses.

The government accepted all 44 recommendations on 6 October.¹ A full roadmap of timelines is due by spring 2027. As of October 2026, then, the UK has a commitment and no rule.

Texas: disclosure tied to diagnosis

Texas has a statute. Senate Bill 1188 took effect on 1 September 2025.³ Its section on AI lets a health care practitioner use AI "for diagnostic purposes," including recommendations on a diagnosis or course of treatment.

That use comes with conditions. The practitioner must act within their licence and review every AI-created record under Texas Medical Board standards. And under subsection (b), a practitioner who uses AI this way "must disclose the practitioner's use of that technology to the practitioner's patients."

The law does not say how or when the disclosure must happen. It also covers only diagnostic and treatment use, not scheduling or billing tools.

California: a label on AI-written messages

California's law is narrower and more specific. AB 3030 took effect on 1 January 2025.⁴ It applies to a health facility, clinic or doctor's office that uses generative AI to write messages "pertaining to patient clinical information."

Those messages must carry a disclaimer that AI generated them. They must also tell the patient how to reach a human health care provider.

There is a large exemption. The rule does not apply to a message "read and reviewed by a human licensed or certified health care provider." So a doctor who checks an AI draft before sending it does not need to label it. Breaches by doctors go to the state's medical boards.

Illinois: consent before AI records a therapy session

Illinois went furthest for one field. Its Wellness and Oversight for Psychological Resources Act took effect on 4 August 2025, according to a summary by the law firm Baker Botts.⁵

Under the act, a licensed therapist may use AI for administrative and some supporting tasks. AI may not make independent therapeutic decisions or talk to clients directly. Before AI records or transcribes a session, the therapist must disclose it in writing and get the client's written, revocable consent. The state can fine up to $10,000 per violation.

The EU: a chatbot must say it is a chatbot

The EU's AI Act added transparency duties on 2 August 2026.⁶ Under Article 50, people must be told they are "interacting with an AI system" unless that is obvious from the context. AI-generated images, audio and video must also be marked as such.

That covers a clinic's AI chatbot or an AI voice on the phone. It does not, on its own terms, require a doctor to say that software helped read your scan. Article 50 is about who or what you are talking to, and about synthetic content.

South Korea: notice now, fines later

South Korea's AI Basic Act took effect on 22 January 2026.⁷ Law firms Cooley and Linklaters both describe the same core duty.⁸ A business must give users advance notice when its product or service uses high-impact AI or generative AI. Both firms list healthcare among the high-impact sectors.

The obligations fall on companies that develop or use AI in their services, which Cooley notes does not obviously exclude hospitals.⁷ Fines run up to 30 million won, about $21,000 by Cooley's conversion. They start only after a one-year grace period, so from early 2027.

Korea matters here because it is a destination. A foreign patient at a Seoul clinic is under Korean rules, and those rules include this notice duty, even if enforcement is a year away.

Turkey: no AI law in force

For patients heading to Turkey, the picture is different. As of the most recent update of the law firm CMS's AI regulation guide, Turkey has no AI-specific law in force.⁹ A draft AI law from June 2024 remains under committee review.

CMS reports that Turkey's personal data law, Law No. 6698, requires transparency where AI processes personal data. It also notes that the data authority's guidance on generative AI is non-binding. None of this amounts to a duty to tell a patient that AI read their scan. This status comes from a law firm's summary, not from a Turkish ministry, and it could change if the draft passes.

What disclosure does not tell you

It is not a quality check. Every rule above concerns whether you are told. None tells you how accurate the AI was, or what it got wrong.

Most rules cover only some uses. California labels AI-written messages, Texas covers diagnosis and Illinois covers therapy. A scan read by AI in California, or a billing tool in Texas, may need no notice at all.

Home rules do not travel. Each rule binds providers in its own jurisdiction. A Californian treated in Istanbul is covered by Turkish rules, not by AB 3030.

The US baseline is still thin. Sara Gerke, a University of Illinois researcher, spoke to Medical Economics in October 2025.¹⁰ She said US informed-consent doctrine "does not necessarily impose a duty to disclose the use of AI in most cases." She sees exceptions when a patient asks, or when AI plays a material role in a treatment decision.

How to ask, wherever you are treated

Put the question in writing before you travel: will AI be used to read my scans, draft my treatment plan or answer my messages? Ask which product, and whether a named doctor reviews its output.

Ask for the answer to go in your records. Where no law requires disclosure, a written answer is the only record you will have.

If a clinic uses an AI chatbot or voice line, ask to speak to a person. In the EU and Korea, rules already require that you be told when you are dealing with an AI system, so asking costs nothing.

Footnotes

1. https://www.gov.uk/government/news/government-backs-recommendations-of-nhs-doctors-led-ai-commission — UK government, "Government backs recommendations of NHS doctors-led AI commission" (published Oct 6, 2026). Accessed Oct 12, 2026. ↩ ↩² ↩³

2. https://assets.publishing.service.gov.uk/media/6aa2662962ec7fe7bedf52b9/Full_National_Commission_Report_-_Publication__1_.pdf — National Commission into the Regulation of AI in Healthcare, "Recommendations for a future regulatory framework," pages 79-80, citing Health Foundation deliberative research (published Sep 10, 2026). Accessed Oct 12, 2026. ↩ ↩² ↩³

3. https://capitol.texas.gov/tlodocs/89R/billtext/pdf/SB01188F.pdf — Texas Legislature, S.B. No. 1188, enrolled text, Sec. 183.005 (effective Sep 1, 2025). Accessed Oct 12, 2026. ↩ ↩²

4. https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202320240AB3030 — California Legislature, AB-3030 "Health care services: artificial intelligence," Chapter 848, Statutes of 2024 (approved Sep 28, 2024); read via Sidley Austin's copy of the chaptered text. Accessed Oct 12, 2026. ↩ ↩²

5. https://ourtake.bakerbotts.com/post/102lr2p/illinois-enacts-first-in-the-nation-guardrails-on-ai-therapy — Baker Botts, "Illinois Enacts First-in-the-Nation Guardrails on AI Therapy" (published Aug 2025). Accessed Oct 12, 2026. ↩ ↩²

6. https://artificialintelligenceact.eu/article/50/ — Future of Life Institute, EU AI Act Explorer, "Article 50: Transparency Obligations for Providers and Deployers of Certain AI Systems." Accessed Oct 12, 2026. ↩ ↩²

7. https://www.cooley.com/news/insight/2026/2026-01-27-south-koreas-ai-basic-act-overview-and-key-takeaways — Cooley, "South Korea's AI Basic Act: Overview and Key Takeaways" (published Jan 27, 2026). Accessed Oct 12, 2026. ↩ ↩² ↩³

8. https://techinsights.linklaters.com/post/102js56/koreas-wonsouth-koreas-ai-basic-act-asias-first-comprehensive-ai-legislatio — Linklaters, "South Korea's AI Basic Act: Asia's first comprehensive AI legislation." Accessed Oct 12, 2026. ↩ ↩²

9. https://cms.law/en/int/expert-guides/ai-regulation-scanner/tuerkiye — CMS, "AI laws and regulation in Türkiye," AI Regulation Scanner. Accessed Oct 12, 2026. ↩ ↩²

10. https://www.medicaleconomics.com/view/inside-ai-malpractice-law-should-physicians-disclose-when-ai-is-used-in-care- — Medical Economics, "Inside AI malpractice law: Should physicians disclose when AI is used in care?" (published Oct 24, 2025). Accessed Oct 12, 2026. ↩

Show all 10 sources