Kaiser says AI isn’t making mental health decisions. A new complaint says otherwise 65%
By Jennifer Wadsworth23%
7/21/2026, 1:00:00 PM
BS Summary: This article contains 37 faulty reasoning types, including Hasty Generalization, Anecdotal, and Optimism Bias, with Negativity Bias as the most egregious example at 33.1% saturation with 346 hits. Analysis detected 2,499 faulty-reasoning hits from 1,046 analyzed words, generating a BS Score of 59.2% and a BS Rank of 65% (7,726 of 21,887 articles). This article is worse (more manipulative) than 64.70% of the article peer group.
Ilana Marcucci-Morris spends less time doing triage these days and more time, as she describes it, cleaning up after robots.
As one of the licensed clinicians Kaiser Permanente patients reach when they call seeking mental healthcare, she used to spend 10 to 15 minutes on the phone assessing for evidence of suicide risk, domestic violence, substance use, or homicidal thoughts before deciding what kind of care was needed.
Now that initial screenings are often handled by phone clerks or an online questionnaire, she increasingly finds herself fixing mistakes made by operators and automated tools.
“I call it ‘service recovery’ more than official triage,” she said.
Subtle cues she picks up during assessments — tics, tone, even words left unsaid — are being missed by automated screenings and untrained clerical staff.
Patients are more likely to arrive at their first appointment intoxicated, in withdrawal, or confused about why they’re seeing a social worker instead of a psychiatrist who can prescribe medication.
Others are referred to outside telehealth providers who later decide they’re not equipped to treat them, sending them back to Kaiser weeks later, angrier and worse off.
Still others wait weeks for care that once took days.
When people reach out for mental healthcare, that first step can be one of the most crucial to get right.
“Our job at the beginning is to instill hope,” Marcucci-Morris said.
“You’re not well now, but I want to give you hope that you can get well.”
Delays and mistakes at that tenuous early stage risk “reinforcing hopelessness” instead.
“Even if other things don’t go well in behavioral healthcare,” she added, “most science says you better nail it at the start.”
At the center of the dispute is a simple question: When someone seeks mental healthcare, who should decide what happens next?
Kaiser therapists say the system is increasingly allowing software and unlicensed staff to make screening decisions that California law says must be made by licensed clinicians — an allegation at the heart of a new regulatory complaint and a hearing this week at San Francisco City Hall.
A test case for AI in healthcare
The dispute unfolding at Kaiser, California’s largest private employer, as it negotiates a contract with the National Union of Healthcare Workers has been described as one of the nation’s first labor fights over AI.
So it’s fitting that in San Francisco — the global hub of artificial intelligence — a local lawmaker is convening a special hearing Tuesday about contract demands that would let the HMO expand the use of automated tools in mental healthcare and lay off the therapists who currently provide it.
Supervisor Chyanne Chen, who called for the hearing, said she wants to make sure ongoing negotiations between Kaiser and the NUHW protect patients as much as employees.
“Every single person, their story is a little different, their circumstances are different, their needs are very different,” she said.
That complexity, she added, doesn’t lend itself to the standardization of an algorithm.
Kaiser has repeatedly denied using AI to replace human care.
“Artificial intelligence holds significant potential to benefit healthcare by supporting better diagnostics, enhancing patient-clinician relationships, optimizing clinicians’ time, and ensuring fairness in care experiences and health outcomes by addressing individual needs,” the company told reporters when its therapists went on strike in the spring.
But a complaint from NUHW offers new evidence that the union says shows Kaiser is already leaning on automation to make the kind of clinical calls that are supposed to be handled by licensed providers.
The complaint, filed Monday with California’s Department of Managed Health Care and the U.S.
Department of Labor, centers on Kaiser’s e-visit tool, an online questionnaire the company directs patients toward when they’re seeking care for anxiety or depression.
To test the tool, the union had four Kaiser members complete it, answering the questions to reflect various levels of symptom severity.
Each time, the tool generated a care recommendation — a referral to virtual therapy, a referral to an outside provider, or simply a suggestion to try a self-care app — within two seconds of the last question being answered, according to screenshots included in the complaint.
That’s hard to square with how Kaiser describes the screening to patients.
In a December 2025 post on its website, the company said that when a member completes an e-visit, “the answers are reviewed by a healthcare professional, and the member receives a response by secure message within four hours, and usually sooner.”
For Marcucci-Morris, the instant turnarounds uncovered by the union testing confirmed what she’d suspected for years.
“That’s 100% of my job,” she said.
“But now it’s automated or algorithmic or AI or whatever it is, making decisions that have taken me decades to be able to legally do.”
A black box
Kaiser has never disclosed how the e-visit tool works, despite repeated requests, according to Fred Seavey, NUHW’s research director.
That lack of transparency troubles Marcucci-Morris most.
If she makes the wrong call, there’s a clear line of accountability.
“If I commit malpractice, the state can take my license away from me,” she said.
“What happens when technology gets it wrong?
Who is accountable?”
Kaiser’s mental health system has faced repeated regulatory action.
In 2023, it agreed to a $200 million settlement over failures to provide timely care.
A subsequent state survey found that it still could not demonstrate adequate suicide-risk screening, and in February it agreed to pay $31.1 million to settle a federal investigation into its use of questionnaires to restrict access to care.
To Seavey, that history undercuts any assurance Kaiser might offer about the algorithm behind its e-visit tool.
“You’ve had to pay out nearly a quarter billion dollars for dozens of violations, and you’re asking us to trust you?”
he said.
“We want to see it.
Why can’t you show us the algorithm?”
Kaiser asked for more time to respond to questions about the complaint.
Tuesday’s hearing, Chen said, is a chance to get answers before the next contract is signed.
Marcucci-Morris plans to be there.
“I have never felt better because a chatbot gave me a reassuring statement,” Marcucci-Morris said.
“We heal through empathy.
That is not something technology can give people.”
Speakers
4speakers48%attributed speech547writer words
Voice mapSelect a segment to jump to its words
Selected voice
100%flagged-word coverageKaiser
115 attributed words23% of attributed speech97% writer coverage
Attribution is sentence-level. Pattern percentages are calculated only from words assigned to that voice.
Loading…
Loading…
Loading…
Analysis
Hover over highlighted words in the article to view the associated bias or fallacy analysis.