The Future of Medical AI — A 5-to-10-Year Outlook and Its Principles— Technology trends, patient-centeredness, the evolution of regulation, and the role of people

01Technology Trends — What Grows for Certain, and What Remains Unknown

Start with the technology outlook. The important thing here is to separate "the parts that will grow for certain" from "the parts running ahead of the evidence." Blur the two, and both investment and field adoption go wrong.

High-confidence is the practical arrival of multimodal AI (= AI that handles several kinds of data — text, images, numbers — together). Medicine is the work of cross-checking different kinds of information — images, lab values, chart notes, genomics — to reach a judgment. The infrastructure to handle these as one is falling into place, and image-diagnosis support and document-drafting support are already entering the field. Still largely unknown, by contrast, is the "delegation of judgment," such as autonomously deciding a treatment plan. Even as the technology draws near, the assignment of responsibility and the mechanisms for verification have not caught up.

The common thread is that "support" comes fast, "delegation" comes slow. Uses that help human judgment take root broadly within a few years; uses that replace human judgment require the design of responsibility and verification before the technology, not after. So when you read a technology trend, ask first "is this support, or delegation?" and you will rarely be wrong.

02The Patient-Centered Axis — Whom Is the Technology For

Talk about technology and the eye naturally drifts to "what it can do." But the axis you must never lose in the future of medical AI is patient-centeredness (= placing the purpose of the technology in the patient's benefit). Being able to process faster, cheaper, and at greater volume is not, in itself, value for the patient.

In its 2021 guidance, the World Health Organization (WHO) placed "protecting human autonomy and well-being" at the head of the principles medical AI must uphold. People first, not efficiency. This order is a foundation that should hold even five to ten years out. Concretely, it means keeping certain questions perpetually turned on yourself — does this AI deepen the patient's understanding, or leave the patient behind? Is the data used for the patient's own benefit? Can the reason for a judgment be explained to the patient?

Brought down to the pharmaceutical field, when you use AI to build patient-facing explanatory materials or disease-awareness content, the crux is that "gets good engagement" and "serves the patient" are different things. Stoking anxiety may lift page views, but that is the opposite of patient-centered. The faster the technology, the more deliberately you have to re-grip this axis.

03The Evolution of Regulation — From Chasing to Designing Ahead

Regulation is often thought of as "something that lags and chases the technology," but the current has shifted over the past few years. Countries are moving toward deciding the frame first, then using medical AI. In Japan too, while the foundation of the Pharmaceutical and Medical Devices Act (= the law governing the quality and advertising of drugs and medical devices) stays unchanged, the treatment of AI-embedded program medical devices has been organized.

Here, let us reconfirm the foundation that anyone involved in pharmaceutical advertising and information provision must never get wrong. Under the Pharmaceutical and Medical Devices Act, the prohibition of exaggerated advertising is Article 66, the prohibition of advertising unapproved drugs is Article 68, and information provision for proper use is Article 68-2. Even as AI makes materials faster to produce, these provisions do not move. In addition, the Sales Information Provision Activity Guideline (= HanteiG, the Guideline on Sales Information Provision Activities for Prescription Drugs, a notice from the Director-General of the MHLW Pharmaceutical Safety and Environmental Health Bureau, 2018), the practical standard for information provision on prescription drugs, and the Standards for Proper Advertising of Drugs and the Like (= the yardstick for advertising expression set out as a notice from the Director of the Compliance and Narcotics Division, Pharmaceutical Safety and Environmental Health Bureau, MHLW) do not slacken depending on whether the maker is AI or human. The evolution of regulation means that these foundations are kept intact while the operational details are updated to fit the new tools.

04The Role of People — The Better AI Gets, the More the Work Shifts

You often see the framing "when AI gets smart, human jobs disappear," but what is happening in the medical field is not disappearance — it is a shift. The more AI takes on routine processing, the more the human role moves from "the one who does the work" to "the one who checks, decides, and explains."

Material review makes this easy to see. As AI handles drafting and first-pass checks quickly, the reviewer's job shifts from "writing the wording from scratch" toward the judgments a machine cannot be left with — "will this expression reach the physician without misunderstanding?" "does this evidence truly match the primary source?" Leave simple deviation detection to the system, and let people handle the subtle judgment. This division of labor becomes the standard going forward.