Waleed Mohsen Warns That 98% AI Accuracy Still Risks Patient Safety

By Waleed Mohsen

San Francisco, CA | September 6, 2025

Artificial intelligence is being deployed rapidly in healthcare, but Waleed Mohsen warns that even a 98 percent accuracy rate can put patients at risk. In a recent commentary, Mohsen explained that small error margins in clinical documentation and decision support can lead to compliance failures, financial losses, and potential harm to patients.

“People often hear 98 percent and assume the technology is safe,” said Waleed Mohsen. “But in healthcare, that missing two percent can mean a suicide risk assessment that was skipped, a safety plan left incomplete, or a billing compliance failure that jeopardizes an organization’s accreditation.”

Independent watchdogs back up this concern. The ECRI Institute, which publishes an annual list of patient safety hazards, has repeatedly flagged improper or incomplete use of AI and clinical decision support systems as a top risk in healthcare. A compliance miss in even a small percentage of encounters can multiply across thousands of patient interactions, leading to preventable harm and millions in denied claims or penalties.

Mohsen emphasized the need for AI oversight. “Trust in technology is not enough. Healthcare organizations must treat compliance as the independent layer of verification, monitoring every interaction whether it involves humans or AI agents.”

The growing use of AI scribes, chatbots, and clinical copilots makes this challenge more urgent. Healthcare providers and payers are now expected to demonstrate not only quality of care but also quality of documentation and adherence to protocols.

For more on how AI compliance can act as healthcare’s new air traffic control system, visit the About Waleed Mohsen page.