A former director who oversaw research at the Mayo Clinic accuses the healthcare system of retaliation after she raised the alarm about “fundamental lapses” in evaluating AI tools for patient care and research.
In a July 6 lawsuit filed in federal court, Traci Tamiko Eto, MA, claims that her insistence on “adhering to federal and institutional standards regarding AI governance and the IRB [internal review board] process” led to her firing.
The legal battle highlights the crucial oversight role that medical institutions play in evaluating whether various types of AI software are appropriate to be used for clinical and research purposes.
“The lawsuit touches on issues that are top of mind for clinicians,” said Taylor N. Anderson, MD, clinical informatics fellow at Oregon Health & Science University in Portland, Oregon, who isn’t involved in the case. “How do I know how much can I trust AI? Has it gone through the proper procedures? What is my responsibility in implementing it in my clinical practice?”
‘Wants Mayo to Do Better’
“This case is about an institution, one of the renowned medical institutions in the world, that decided being first and being fastest in AI is more important than following the rules and preserving the integrity of your process,” Eto’s attorney Artur Davis, JD, of Birmingham, Alabama, told Medscape Medical News. “She wants Mayo to do better, and she wants Mayo to live up to the standards that it claims it adheres to.”
Mayo Clinic declined to comment on the lawsuit but said in a statement it “is committed to the responsible development and deployment of AI, with privacy, security, transparency, and compliance embedded throughout our processes. Our research and clinical innovation are conducted in accordance with applicable laws and regulations, and we remain steadfast in upholding the trust patients place in us and respecting their privacy.”
Inside Lawsuit’s Allegations
According to the lawsuit, Eto was hired by Mayo Clinic in December 2023 after previously working on compliance with research standards at Stanford Research Institute International, Kaiser Permanente, and Chicago’s Lurie Children’s Hospital. Eto oversaw the Mayo Clinic’s Human Research Protection Program and its IRB, managing 36 workers and 3 managers.
At Mayo, Eto alleges that she filed complaints about irregularities such as:
- The bypassing of IRB review in several instances including the de-identification of shared data and the sale of patient biospecimens to a commercial entity.
- Pressure on staff to approve informed consent waivers or forgo documenting them.
- The authorization of “a high-risk investigational medical device to perform cardiac surgery although the procedure had never obtained IRB review.”
According to attorney Davis, a former US congressman, the claims represent “a 1-year pattern of Mayo consistently losing its way and almost being aggressive about it. Taking the IRB panels and distorting them or manipulating them, being aggressive about pushing members of her team to steer toward particular results, or waiting to get panels that they thought had a favorable composition.”
AI-Related Allegations
In regard to AI, the lawsuit alleges that Mayo Clinic staff exempted a study from IRB review that aimed to analyze Mayo’s digital assistant tool, known as MAYA.
Researchers “mischaracterized a series of outcomes, deleted unfavorable results,” and deployed “unsanctioned” software without regulatory authorization, according to the lawsuit. This “compromised both safety and data security because it injected the software into ordinary clinical workflow with which MAYA was interacting.”
The Mayo Clinic says more than 200 AI projects are in “various stages of maturity,” and a June 2026 post says more than 100 clinical applications have been reviewed by staff this year alone.
In a video posted by Mayo Clinic on Instagram, a professor of biomedical informatics says: “In order to design any type of AI tools, they need to be centered by the needs of our patients and clinicians. And not only that, we also emphasize a lot on AI regulation and governance. For each AI tool, we need to ensure safety and trustworthiness before it can actually be used into patient care.”
Mayo ‘Breaking Promise It Signed’
The lawsuit alleges that a manager told Eto in March 2025, after she had filed her complaints, that she needed to either quit her research director job or face being “unemployable” at Mayo.
Eto claims that she developed depression and anxiety, was approved to go on leave only after initially being rejected, and learned in September 2025 that her position was being eliminated. She applied for 15 internal jobs but was not interviewed and was finally terminated in December 2025.
The lawsuit claims that Mayo Clinic violated the False Claims Act, which Davis said protects employees who act to stop violations of federal grant certifications. “Every time the review process was not followed in the right way, Mayo was breaking the promise that it signed,” he said.
The lawsuit says that since leaving Mayo Clinic, “Eto’s career prospects have been moribund, with no viable prospects.” However, LinkedIn lists her as a senior teaching fellow at the Center for AI and Digital Policy, and a research consulting firm announced her hiring in January 2026.
Recommendations for Using AI in Medicine
How should medical professionals use AI software if they’re not sure it’s been properly vetted by their health systems?
“Clinicians’ responsibilities lie within their own clinical practice and judgment,” said Michael H. Bernstein, PhD, associate professor at Warren Alpert Medical School of Brown University, Providence, Rhode Island, who studies AI’s use in radiology.
“I would not advocate that they should bear responsibility for the AI systems themselves,” he said. “This is because they generally lack the technical expertise to evaluate AI. That requires advanced training in statistics, which exceeds what is provided in medical education. Rather, physicians’ evaluations of AI will be limited to a more impressionistic perspective, such as its perceived impact on their own clinical judgment.”
Bernstein’s colleague Grayson Baird, PhD, associate professor and vice chair who also studies the use of AI in radiology, told Medscape Medical News that it’s probably best to assume that every AI-assisted clinical output could be a false positive or false negative.
“In addition,” he said, “our research indicates that radiologists should first review and interpret imaging before receiving AI output and document both. Our research suggests that if a radiologist believes an abnormality is present before seeing the AI output, it is likely best to go with their initial impression and not review the AI output.”
Baird and Bernstein reported having no disclosures.
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.