​

Editor's Note

​

What AI Owes to Humanity — And the Risk of Failing to Deliver

​

By Dr. Qaisar J. Qayyum, Chief Editor, Noor Journal

​

Medical artificial intelligence is becoming embedded in healthcare systems globally. By mid-2026, major EHR vendors are releasing AI documentation tools as default features. By late 2026, AI-generated clinical notes will be accepted by insurance companies for billing purposes. This is not speculation, it is the announced timeline of Epic, Cerner, and Oracle. The technology is already deployed. The scale is accelerating.

​

Both physicians and patients will increasingly rely on these systems for clinical guidance. Neither group fully understands what they are trusting. Both will act on what the systems tell them.

​

Here is what is happening right now, not hypothetically. A 62-year-old woman opens a consumer AI health app and searches "should I take aspirin?" The system responds that aspirin reduces cardiovascular events by "25%." Her physician, consulting UpToDate Expert AI for the same patient, receives the identical figure. Both patient and physician see the same number. Neither sees that this number is relative, not absolute, that the actual benefit is 35 events prevented per 1,000 patients over two years. Neither sees the numbers for bleeding risk. Patient may neglect to mention that her chronic kidney disease and prior GI bleeding, which change the calculation entirely. Neither sees that ACC/AHA and USPSTF guidelines contradict each other on whether she should take it at all. Neither sees the industry funding behind those guidelines.

​

The patient makes a decision. The physician makes a recommendation. Both are based on incomplete information presented in the language of authority.

This is not theoretical. Nine million Americans over age 70 are currently taking aspirin for primary prevention despite guideline reversal. These are real patients. Real decisions made on incomplete data. Real potential for harm.

​

The question is not whether medical AI will become standard. The question is whether it will become standard with transparency or without it. The window for answering that question is closing. Once AI is embedded as the default in every major healthcare system, changing it becomes economically and politically prohibitive. The incomplete framing becomes structural. The bias becomes invisible because it is universal.

​

We cannot prevent this development. We can only shape it. The World Health Organization has the standing to establish standards that might require: absolute risk figures presented before relative ones, baseline event rates disclosed for every claim, natural frequency equivalents for every efficacy statement, explicit disclosure of conflicts of interest, identification of guideline sources and funding, side-by-side presentation of contradicting guidelines, harm data in identical format as benefit data.

​

National regulatory bodies have the authority to classify clinical AI as a medical device requiring approval before deployment, ensuring systems meet standards before reaching patients and physicians.

​

Medical societies have the professional credibility to audit AI outputs, identify failures, and coordinate globally to prevent regulatory arbitrage.

These three mechanisms must work together. WHO sets standards. Regulators enforce them. Societies monitor compliance. Each without the others fails. Together, they could embed transparency into medical AI before it becomes universal.

​

One day, we will all be patients. One day, we will all consult machines. One day, we will hope that someone fought to ensure those machines tell us the truth, not the profitable truth, not the convenient truth, but the complete truth that allows us to choose for ourselves.

​

The Noor Journal believes this standard is worth pursuing now, while pursuing it is still possible.

​

Dr. Qaisar J. Qayyum, MD

Chief Editor, Noor Journal of Complementary and Contemporary Medicine

Clinical Assistant Professor, Internal Medicine and Geriatric Medicine Oklahoma, USA

​