A former research director at Mayo Clinic is alleging that the health network concealed error rates of up to 67% in its flagship AI digital assistant, MAYA. If factual, these claims should shock the entire healthcare community: an AI tool with a 67% error rate does not function.
The purported cover up is even more upsetting. The AI leader, Traci Tamiko Eto, joined Mayo in December 2023 to align the health system’s research practices with federal standards. Eto now claims that Mayo terminated her employment after she consistently flagged system failures and shoddy research practices. According to the lawsuit, over Eto’s 18 months at the health system, she flagged mishandled patient data, cardiac surgical devices approved without institutional review, and consistent failure to comply with executive research standards. After Eto and 10 separate whistleblowers reported MAYA’s failure rate to Mayo executives, she alleges that her supervisor reprimanded her for "jeopardizing the pace of ongoing research projects” and “compromising Mayo’s competitive advantage.”
Eto elaborates that after she continued to raise compliance concerns, executives excluded her from leadership meetings, demoted her, and then finally terminated her employment. If true, Mayo's termination order is ostensibly illegal under the False Claims Act, which shields whistleblowers from employer retaliation. Perhaps setting up future lawsuits under the False Claims Act, Eto also alleges that Mayo maintains a “ghost file” which marks employees who reported research misconduct as “not eligible for rehire.”
Consider a paradigm oft-repeated by CHARGE: we only learn about AI system failure through forced disclosure. So far, we’ve applied this analysis in the context of vendor failures, as in Xsolis, Sentri7, and nH predict. What makes this case so astounding is that one would anticipate that Mayo Clinic, as a provider, not a vendor, would be the most dedicated to AI compliance and patient safety. For all providers, CHARGE advocates for internal compliance auditing within hospital networks. According to the lawsuit, at Mayo, compliance auditors like Eto accurately flag their concerns. If Mayo Clinic dismisses system failure – CHARGE stresses that the lawsuit is still awaiting adjudication – it is because executives actively and deliberately obscure error rates even with robust compliance institutions in place.
The reminder is that AI governance is as much dependent on mutual trust between healthcare professionals as it is about institutional mechanisms for auditing or reporting concerns. Patient-facing AI tools should be deployed at the point-of-care to deliver improved health outcomes, better patient experience, and alleviate pressure on physicians. The challenge for professionals is interpersonal and, as this lawsuit exposes, discursive: conflating financial advantage with health outcomes from cardiac-surgery devices is an intolerable paradigm that professionals must unite to prevent from replicating across health systems; deliberating ignoring patient risk to maximize profits is ethical impropriety.
Mayo Clinic responded to Becker’s initial request for comment:
Mayo Clinic is committed to the responsible development and deployment of AI, with privacy, security, transparency and compliane 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.
Industry professionals and CHARGE researchers await the outcome of litigation, and CHARGE reporting is dedicated solely to the substance of the lawsuit. This article is not an analysis of the veracity of Eto’s accusations; that obligation is reserved to the U.S. District Court in Minnesota.
References
[2] https://www.mprnews.org/story/2026/07/09/lawsuit-alleges-mayo-clinic-cut-corners-with-ai