Your Therapist's New Silent Partner

Molly Quinn is a 31-year-old librarian from Fayetteville, Arkansas. She spent two years trusting her therapist with things she'd never told anyone else.

When her therapist mentioned trying an AI note-taking tool, Quinn didn't say no outright. She said she wanted to research it first. She wanted to know where her words would go.

According to NPR, the session moved on — and halfway through, Quinn noticed the therapist's iPad was just propped up, pointed at her. No note-taking. Just recording.

She froze. She kept talking. And when she walked out, it hit her.

"This person who I'm supposed to be able to trust with some very private and very intense emotions had just completely disregarded something I said I was not comfortable with," Quinn told NPR. "I felt completely violated."

This is occurring more frequently across the industry.

The Product Being Sold

A growing number of companies are marketing AI tools directly to therapists. One of them is called Berries.

Berries CEO Tal Salman told NPR the platform is "designed to reduce administrative burden without interfering with the therapeutic experience itself." When activated, it records the session, transcribes it, and generates a draft clinical note for the therapist to review, edit, or discard.

The pitch is simple: therapists spend enormous time on paperwork. AI handles the paperwork. Therapists get more face time with clients. Everybody wins.

On paper, that sounds reasonable. In practice, the consent process is where it falls apart.

The Consent Problem Nobody Wants to Talk About

HIPAA — the federal health privacy law — requires that patients be informed when their sessions are recorded. But "informed" carries significant weight in that sentence.

A buried checkbox in an intake form is not meaningful consent. A therapist casually mentioning a new tool mid-session and then propping up a tablet is not meaningful consent.

Psychology.org spoke to a panel of four licensed therapists about ethical AI integration. Licensed clinical social worker Shari B. Kaplan, who has over 20 years of clinical experience and founded Cannectd Wellness, said client safety and informed consent must remain "paramount." She flagged data security as a top concern, noting that "vague policies and past breaches have undermined trust in digital technology."

Vague policies mean you often have no idea what's happening to your data.

What the Data Actually Shows

A 2024 study published in JMIR Mental Health by researchers at Orygen Digital and the University of Melbourne — including Shane Cross, Imogen Bell, Jennifer Nicholas, and Mario Alvarez-Jimenez — surveyed both community members and mental health professionals about AI use in mental health care.

The research makes clear that public comfort with AI in clinical settings is not settled. There is a real and measurable gap between what professionals see as useful and what patients are willing to accept.

A separate 2024 Pew Research Center survey found that 30% of U.S. adults interact with AI at least several days a week. People are familiar with AI. That doesn't mean they're fine with it sitting in on their trauma disclosures.

What Mainstream Coverage Is Getting Wrong

NPR's coverage is solid on the patient experience but soft on accountability. The story centers on Quinn's feelings — legitimate and important — but it doesn't press hard enough on the regulatory gap or name which specific HIPAA provisions are being tested.

The broader media framing treats this as a "both sides have a point" story about convenience vs. privacy. There is no version of secretly recording a therapy patient — after they expressed hesitation — that is defensible.

The left-leaning coverage also tends to frame this primarily as a tech-sector problem, something to be solved with more regulation. But the therapist in Quinn's case is the one who made the call to start recording. The human failed the ethical test before the AI got anywhere near it.

The Real Risks Nobody Is Quantifying

Therapy transcripts contain abuse history, addiction, suicidal ideation, affairs, criminal confessions, and mental health diagnoses that can affect employment, custody battles, and insurance.

Now ask: where does that transcript go? Who can subpoena it? What happens if the AI company gets breached? What happens when the company gets acquired?

Kibby McMahon, Ph.D., a licensed psychologist and co-founder of digital mental health company KulaMind, told Psychology.org that the therapeutic relationship itself is at stake — not just data hygiene.

She's right. And the stakes are higher than a leaked email or a hacked credit card number.

The Stakes

AI note-taking in therapy isn't inherently wrong. Burnout among mental health providers is real. Administrative load is crushing practices. Tools that reduce that burden can mean more therapists staying in the field, which means more patients getting care.

But "useful" and "ethical" are not the same word.

Right now, there is no federal standard specifically governing AI transcription in clinical mental health settings. Regulators are still debating who is even responsible for oversight — government or the tech companies themselves, according to Psychology.org.

In that vacuum, individual therapists are making judgment calls. Some are making bad ones.

Molly Quinn trusted her therapist with two years of her life's most painful material. Her therapist responded by hitting record without permission.

If that's the future of mental health care, fewer people will seek it. And that's a public health disaster waiting to happen.