The AI scribe liability moved to consent. Here is your script.

What the lawsuit actually claims

On April 8, 2026, a group of patients filed a proposed class action in the US District Court for the Northern District of California. The core allegation is specific: that Sutter Health and MemorialCare deployed Abridge's ambient AI scribe to record and transcribe patient visits without obtaining patient consent first.

The suit asserts violations of three laws: the California Invasion of Privacy Act, the Confidentiality of Medical Information Act, and the Federal Wiretap Act. Through May and June 2026, follow-on suits named additional health systems on similar theories.

It is worth reading what this is and is not about. The complaint is not, at its center, an argument that the AI got the note wrong. It is an argument that the room was recorded without permission. That distinction is the whole story for a working clinician, because it changes where the risk sits and what you have to do about it.

Why the risk moved from accuracy to consent

For the last few years, the conversation about ambient AI scribes has been about quality. Does it hear correctly. Does it hallucinate. Does the note need heavy editing. Those are real questions, and they have a separate home, which we will point you to below.

This case is about something else entirely. An ambient scribe works by listening to and recording a conversation. Wiretap and recording-consent laws care about exactly that act: capturing a conversation. In a number of states, recording a conversation lawfully requires the consent of everyone being recorded, not just one party. When a scribe starts listening before anyone has agreed, the recording itself can be the violation, regardless of how good the transcript turns out to be.

That is why the liability moved. A perfect transcript made from an unconsented recording does not fix the problem. The exposure attaches at the moment of capture. For you, that means the protective step is not better editing after the visit. It is consent before the visit is recorded at all.

Why this lands on the clinician, not just the institution

It is tempting to assume this is a hospital problem, settled by whatever the health system signed off on. Do not assume that. The clinician is the person in the room when the recording starts.

A health system can make a procurement decision and write a policy, but consent is an interaction between a patient and the person treating them. If the rollout did not build a real consent step into the visit, the gap shows up at the bedside, and you are the one standing there. The suits naming health systems do not make the individual clinician's exposure disappear, and waiting to find out how that shakes out is not a plan.

The good news is that the fix is entirely within your control and costs you about ten seconds. You do not need the institution to solve this before you protect yourself. You need a consent habit you run every time the scribe is on.

The all-party consent script and documentation step

Here is a practical approach that protects you even while the institutional picture is unsettled. Treat the scribe like any other recording: assume all-party consent is required, get it out loud, and write it down.

A short verbal script, said before you start the scribe:

• "I use an AI assistant that listens to our visit and helps me write the note. It records our conversation to do that. Is it OK with you if I turn it on?"
• Wait for a clear yes from the patient, and from anyone else in the room, such as a family member or interpreter. All-party means everyone present whose voice may be captured.
• If anyone declines, do not start the scribe. Take the visit the old way. One refusal is not a fight to win.

Then the documentation step, which is what turns a good habit into protection you can point to:

• Note in the chart that you obtained verbal consent for AI scribe recording, that the patient agreed, and the date and time.
• If others were present, note that they were informed and consented too.
• Keep the wording consistent so it is obviously a routine you run every time, not a one-off.

That is the entire procedure. Ask, wait for a clear yes from everyone, document it, and only then start recording. It is short enough to do without slowing the visit, and it is the exact thing the consent claims are about.

What this briefing does not cover

This page is strictly about the recording-consent and wiretap angle, because that is the new exposure the April 2026 case put on the table. It is not the place for general HIPAA hygiene or the broader practice of charting safely with AI.

Those questions matter, and they have their own home. For how to use AI for clinical notes safely overall, including the accuracy and privacy practices that sit alongside consent, read How doctors use AI for clinical notes safely. The two pages fit together: that one covers the general safety method, this one covers the specific consent step you add on top of it.

The honest caveat is that recording-consent law varies by state and the litigation is moving. Treat the all-party approach as the safe default, since it protects you in the strictest states, and confirm the specifics for your jurisdiction and your institution's current policy.

The skill under the headline

There will be another lawsuit, another tool, another headline. Reacting to each one as it breaks is exhausting and it leaves gaps. What protects you across all of them is a steady method: knowing where AI helps in your charting, where it creates exposure, and what small habits, like consent before recording, keep you personally covered no matter what the institution does.

That method is what holds up when the news changes. If you want the structured version built for clinicians, AI for Physician Notes teaches it for busy practicing physicians, and the two minute course quiz will point you to the right starting place for your work.