When AI Moves Faster Than Patient Safety

Posted: July 17, 2026
By Howard Farran, DDS, MBA

When AI Moves Faster Than Patient Safety

A federal whistleblower lawsuit against Mayo Clinic is raising a question every dental practice should confront before adopting artificial intelligence. What happens when the pressure to move faster becomes stronger than the discipline to verify what the technology is doing?

Traci Tamiko Eto, Mayo Clinic’s former director of research operations and AI compliance, alleges that she was demoted and later fired after raising concerns about patient privacy, research oversight, and the validation of AI systems. Eto claims that some patient data deidentification processes were not properly reviewed, an investigational cardiac surgery device bypassed institutional review, unfavorable results were removed from a digital assistant project, and investigators attempted to conceal an error rate reported at 67 percent.

These claims remain allegations. They have not been proven in court, and Mayo Clinic denies wrongdoing. The institution says privacy, security, transparency, and regulatory compliance remain central to its AI programs.

The important lesson for dentists is not whether Mayo Clinic is ultimately found liable. It is how easily a respected organization can become vulnerable when innovation, competition, and reputation begin moving faster than governance.

Dentistry is entering the same territory. AI now reads radiographs, drafts clinical notes, records conversations, proposes treatment, predicts insurance reimbursement, writes patient messages, and helps present cases. Each tool promises speed. Speed can improve a practice, but it can also conceal errors before anyone has learned where to look for them.

An AI generated note that confuses hospitalist care with hospice care may sound absurd, but similar errors in dentistry can be quieter and more dangerous. A note may record the wrong tooth, invent a symptom, omit informed consent, misstate the patient’s medical history, or turn a tentative diagnosis into a definitive one. A radiographic system may highlight a lesion that is not present or fail to identify disease that is. A treatment presentation tool may communicate certainty that the dentist never intended.

The dentist remains responsible.

That is the operational reality. AI can assist judgment, but it cannot assume the legal, ethical, or clinical accountability attached to the dentist’s license. Every diagnosis, treatment plan, chart entry, prescription, referral, and patient communication still requires human review.

Privacy presents another risk. Ambient AI scribes can listen to entire patient encounters and convert them into documentation. That can reduce clerical burden, but patients may not understand that their conversation is being recorded, transmitted, processed, or stored by an outside vendor. Consent should not be buried in dense paperwork or assumed because the technology is convenient. Patients should know when AI is listening and how their information is being used.

Practices should also resist the temptation to treat vendor claims as proof. A published accuracy number may come from a controlled dataset that does not resemble the patients in a general dental office. The useful question is not whether the software performed well in a demonstration. It is whether it performs reliably in your operatories, with your images, your accents, your workflows, your assistants, and your patients.

The safest approach is simple. Introduce one tool at a time. Define its purpose. Test it on a limited basis. Track errors. Preserve the original data. Require the dentist to approve every clinical output. Tell patients when AI is involved. Review the vendor’s privacy terms, data retention policies, and responsibility for breaches. Create a clear process for team members to report problems without fear.

The Mayo Clinic lawsuit is often framed as evidence that Silicon Valley’s move fast mentality has entered medicine. That may be partly true, but AI is not the root problem. Deleting unfavorable results, bypassing oversight, concealing limitations, and punishing employees who raise concerns were possible long before machine learning.

AI simply increases the speed, scale, and invisibility of those failures.

A dental practice does not need a formal research department to benefit from the same safeguards used in serious clinical research. Keep records of what the system produced. Compare its output with what actually happened. Investigate repeated errors. Do not allow a vendor, consultant, or enthusiastic team member to dismiss concerns because the technology is new.

The practices that benefit most from AI will not be the ones that adopt everything first. They will be the ones that know exactly where the machine stops and professional judgment begins.

Would you trust an AI system in your operatory if you could not independently verify how often it is wrong?



Join the Conversation!




When AI Moves Faster Than Patient Safety


When AI Moves Faster Than Patient Safety


Primary news reporting

MPR News
Lawsuit alleges Mayo Clinic cuts corners with AI, putting patient care and privacy at risk.
https://www.mprnews.org/story/2026/07/09/lawsuit-alleges-mayo-clinic-cut-corners-with-ai

Inc.
She Tried to Fix Mayo Clinic’s AI. Instead, She Claims Her Job Was Eliminated When She Raised Concerns.
https://www.inc.com/lucia-auerbach/tried-to-fix-mayo-clinic-ai-instead-she-claims-her-job-was-eliminated/91372461

Industry analysis

STAT
AI Prognosis: Mayo whistleblower lawsuit and what it means for AI governance in healthcare.
https://www.statnews.com/2026/07/15/mayo-whistleblower-sutter-abridge-lawsuit-news-ai-prognosis/


Views: 6
Sponsors
Townie Perks