Skip to content
Submit incident
AII-1000-34261
Under investigationA regulator, court or police has opened a formal inquiry and reached no finding yet.

Kaiser Permanente's AI monitoring reduces mental health care quality and surveillance of nurses

Kaiser Permanente deployed AI systems to automate mental health treatment recommendations in seconds and to surveil and grade the performance of advice-line nurses, reportedly compromising clinical judgment and creating incentives that deterred nurses from providing adequate care in crisis situations, including one case involving a suicidal patient.

20 July 2026 Β· Kaiser Permanente Β· Privacy and surveillance
20 July 2026
AII-1000-342614 sources ↓
Quick facts
Category
Privacy and surveillance
Country
United States
Source
CBS News
Date
20 July 2026
Drafted by
Raidu Incident Detection

What happened

Kaiser Permanente used AI systems to make rapid treatment recommendations in mental health screening, with one system generating clinical treatment plans in less than two seconds based on patient questionnaires, and monitored advice line nurses with software that tracks call duration, activity, and predicted productivity.

In 2024, Kaiser piloted software that graded tone and apparent empathy in nurse and patient voices; nurses said the program scored calls incorrectly and petitioned against it, leading Kaiser to suspend the pilot in November, though nurses were told it could return.

Reported impact

Licensed therapists and nurses reported that AI-driven speed pressure and monitoring altered clinical practice: one nurse stayed on the phone with a suicidal patient for over an hour despite knowing it would damage her performance score, and another withheld additional time and compassion from an elderly patient given a terminal cancer diagnosis because she feared score reduction and reprimand.

Nurses received reduced scores for calls exceeding 15 minutes, for departing from software recommendations, or for providing additional advice, and reported insufficient time for calls requiring interpreters, which routinely exceeded 30 minutes, while available post-call time fell from about 10 minutes to 30 seconds or less.

Index analysis

The systems created misaligned incentives: call-time targets conflicted with clinical judgment, particularly in mental health and crisis situations where patient safety required time-intensive assessment and support.

Governance controls

Kaiser disputed claims it uses AI to diagnose patients and said its contact-center tools are subject to human review, but did not address why nurses are questioned about calls exceeding 15 minutes or explain the role of call time in performance scores.

Classification

Category
Privacy and surveillance
Industry
Healthcare, drugs and biotechnology
Country
United States
Harm status
Actual harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Psychological, Public interest, Human or fundamental rights
Harmed
Workers, Consumers
Failure modes
Overreliance, Unsafe autonomous behaviour
Consequences
System suspension or termination
AI systems
Behavioral health e-visit tool, Call monitoring and predictive productivity software, Voice tone and empathy grading software

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

CBS News
Kaiser mental health workers bring contract fight over AI to San Francisco City Hall hearing
24 July 2026
sfstandard.com
Kaiser's mental health AI made decisions in seconds, complaint says
21 July 2026
World Socialist
Kaiser Permanente implementing invasive AI surveillance of advice hotline nurses
21 July 2026
SFist - San Francisco News, Restaurants, Events, & Sports
SF Supervisors Hold Hearing on Kaiser's Push to Expand AI Use, Lay Off Mental Health Staff
24 July 2026