Soviet constructivist graphic: a cracked medical cross entangled in mechanical gears

Mayo Clinic's AI Assistant Was Wrong Two-Thirds of the Time. A Whistleblower Said So and Got Pushed Out Instead.

A former Mayo Clinic research director says staff hid a 67 percent error rate in the hospital's AI assistant and bypassed federal review boards, including for a cardiac surgery procedure. She reported it ten times. She lost her job instead.

An Assistant That Was Wrong Two-Thirds of the Time

Mayo Clinic built an AI-integrated digital assistant called MAYA to help run its research operations. According to a new federal whistleblower lawsuit, staff working on the project knew the tool had an error rate of 67 percent and decided the appropriate response was to keep that number to themselves. The complaint, filed by former research director and AI compliance lead Traci Tamiko Eto, alleges the project team overstated MAYA's capabilities and quietly removed unfavorable test results before anyone outside the room could see them.

Even now, Mayo has not disputed the number. It has simply declined to discuss it.

The Reviews That Never Happened

Federal research rules exist for a reason nobody at Mayo apparently felt like explaining to an Institutional Review Board. Eto's suit alleges that over roughly eighteen months, leadership bypassed IRB review outright in multiple cases, steered proposals toward whichever board was least likely to ask hard questions, and in one instance authorized a cardiac surgical device procedure without the review federal law requires. Sources within the Medical Research Community will tell you, patiently, that this is the entire point of an IRB: to catch the thing nobody wants to catch. In this case, apparently, the catching was optional.

The patient on that cardiac table did not get a vote on which corners were cut.

Eighteen Months, Ten Reports, One Door

Eto did not raise this once and walk away. She filed ten separate whistleblower complaints. Her supervisor, she says, dismissed the privacy concerns and prioritized research pace over compliance, which is a generous way of describing a hospital choosing speed over the mechanism that exists to stop it from hurting people. By early 2025, she was excluded from executive meetings and told, in the specific corporate dialect reserved for people about to be pushed out, that she was a "poor cultural fit." The offer that followed was resignation, or a personnel file rewritten to make the exit look like something other than what it was.

In a development that will surprise no one who has been paying attention, the employee flagging the problem left. The problem, as far as the record shows, did not.

The Institution Responds

Mayo Clinic says its "research and clinical innovation are conducted in accordance with applicable laws and regulations," and has otherwise declined to comment on pending litigation, which is the corporate equivalent of a shrug performed by a legal department. The Mayo Clinic Platform, the AI-integrated data system at the center of Eto's original privacy complaints, is still running. The Algorithm, whatever its actual accuracy, remains fully credentialed.

The lawsuit will take its course. The 67 percent, one presumes, was never actually fixed so much as filed.

Sources: MPR News · Futurism