Healthcare AI Safety Gap · Don't break patients
    Commercial survey / litigation9 Jul 2026·US

    Whistleblower alleges Mayo Clinic hid its AI assistant's 67% error rate

    Former Mayo research director's suit alleges staff knew MAYA erred two-thirds of the time and bypassed IRB review to move faster on AI.

    67%
    error rate allegedly known internally and covered up

    What the research found

    A federal lawsuit from former Mayo Clinic research director and AI compliance lead Traci Tamiko Eto alleges the health system covered up or ignored research results showing its MAYA digital assistant had a 67% error rate, and that leaders pressured or bypassed institutional review boards to speed AI research, in one case greenlighting a high-risk, unvetted AI-enabled device for cardiac surgery abroad over her objection. She says she was sidelined and then pushed toward resignation after repeatedly raising the concerns. Mayo says it does not comment on active litigation but is committed to the responsible development and deployment of AI.

    Why it matters for providers

    A specific, named allegation that a major health system's own staff knew a patient-facing AI tool was wrong two-thirds of the time and deployed it anyway, precisely the quiet, hidden-failure pattern this tracker exists to surface. Claims are unproven allegations in active litigation.

    Original source
    Minnesota Public Radio (reporting on ongoing litigation)
    Read the full report ↗
    A commercial survey or unproven litigation — indicative only; treat with caution.