Litigation Update: AI Transcription Use in Hospital Environments

A proposed class action in the Northern District of California (Washington et al v. Sutter Health et al, No. 4:26-cv-03012) alleges that Sutter Health and MemorialCare recorded patient visits through Abridge's ambient AI scribe without consent. The AI vendor is not a defendant. The hospitals that deployed it are. No breach, hack, or outage is alleged. The claimed harm is the ordinary, intended use of the tool.

Key takeaways for brokers

• The deployer, not the developer, is the named defendant.
• The claims are privacy and consent claims, not a cyber incident. No attack is alleged.
• Exposure scales with rollout. In submissions we have reviewed, some hospitals run hundreds of AI scribe licenses.
• Four questions (below) tell a broker whether a hospital client has the consent and governance evidence it would need to defend a claim like this one.

Why can AI transcription lead to litigation for hospitals?

During our litigation vulnerability analysis, we uncovered a proposed class action against Sutter Health and MemorialCare (Case 4:26-cv-03012) following a recent submission from a hospital concerned about unauthorized data disclosure. While our submission is unrelated to the defendants in this class action, it provides a significant liability risk signal for hospitals in California using Abridge or similar ambient AI scribes.

The complaint alleges that Abridge was used to capture physician-patient conversations, transmit them outside the clinical setting, and process them through third-party systems without adequate notice or informed consent.

These are currently allegations, not yet judicial findings, as this is an ongoing lawsuit. Abridge, the developer of the AI transcription tool, is not yet reported to be a defendant in the action; rather, Sutter Health and MemorialCare, the deployers of these tools, are the named defendants.

This case highlights how deploying AI tools, such as AI scribes, can lead to litigation and potential liability if proper consent is not obtained.

In addition, this complaint is an excellent example of how privacy lawsuits can arise from the use and deployment of an AI tool, even in the absence of an adversarial attack or cyber incident. We made the same point in our Q1 2026 AI liability insurance market update: the cases that reach deployers are rarely security failures.

What allegations could hospitals face over AI scribes?

Hospitals using Abridge are not simply converting speech into text. Using AI scribes can create a chain of activity involving audio capture, external transmission, AI processing, draft clinical documentation, and integration with electronic health records.

If hospital patients are not properly informed, consent is not adequately evidenced, and generated notes are not meaningfully reviewed, a hospital could face allegations involving:

• Unauthorized recording of confidential consultations
• Improper disclosure or processing of protected health information
• Privacy and consumer-protection violations
• Errors in AI-generated clinical records
• Failure to implement system-wide safeguards
• Inadequate AI usage policies/protocols governing physician conduct

In some of the insurance submissions we have reviewed, AI scribes are already being rolled out extensively among clinical professionals. In some cases, we have seen hundreds of licenses per hospital. If there is any oversight or weakness in the consent or disclosure processes, this could create systemic litigation exposure.

What risk management questions should brokers ask?

Four questions a broker can put to a hospital insured. Each is paired with the language from the complaint that makes it matter.

1. How is the Insured deploying AI transcription tools?

1. Who uses it? (licensed physicians, RNs, hospital staff)
2. How do they use it? (AI usage policies, training modules)
3. Has the Insured implemented safety protocols? (guardrails, human review of AI-generated content)

From the filing: ‘The use of the Abridge system was not limited to the encounters involving Plaintiffs. Rather, upon information and belief, the platform was implemented across numerous clinics and provider practices operated or affiliated with Defendants, resulting in the recording of a substantial number of patient visits throughout Defendants’ healthcare networks’ … ‘the recordings generated through the Abridge system were routinely transmitted to external infrastructure associated with the platform for transcription, analysis, and the generation of draft clinical documentation. These recordings and transcripts contained patients’ confidential medical communications, including discussions of symptoms, diagnoses, treatment plans, medications, and other individually identifiable health information’.

2. What do the Insured’s consent and disclosure policies look like?

Do they have internal procedures to ensure that employees disclose their use of AI tools at the outset of a meeting? Must employees obtain patient consent?

From the filing: ‘Among the common questions of law and fact common to the Class are: Whether Defendants had a policy or practice of using Abridge or similar technology to record patient conversations without obtaining all-party consent’ … ‘Whether Defendants failed to implement any system-wide protocol reasonably designed to ensure that patient consent was obtained prior to recording … Whether Defendants sought or obtained prior consent—express or otherwise—from Plaintiffs and the Class’.

3. What does the Insured perceive as the exposure?

Which downstream risks concern them?

From the filing: ‘Defendants implemented and maintained this recording practice without establishing uniform procedures to ensure that patients provided prior, express, and informed consent before their confidential medical conversations were recorded and transmitted through the Abridge system’ … ‘Defendants’ pattern and practice caused class-wide harm, including the invasion of patients’ privacy, loss of control over their confidential medical information, emotional distress, and the increased risk that sensitive medical communications could be accessed, disclosed, or misused by unauthorized persons’.

4. How would the Insured’s existing coverage respond?

With the context of Question 3 considered, how do you (as the broker, or the Insured) anticipate the Insured’s E&O, medmal, and cyber coverage responding to a claim like this? Where the answer is unclear, standalone AI liability insurance is written for exactly this gap.

How can brokers help hospital clients using AI scribes?

Hospitals should not be complacent when deploying AI scribes if they transcribe and generate outputs containing confidential medical information. The failure to obtain consent and the omission of AI governance could lead to litigation.

If you are an insurance broker, now is a great time to get ahead and help your hospital clients who are using AI transcription tools such as Abridge. Using just a company’s name and website, we can generate intelligence reports for brokers to help you identify whether any clients in your portfolio have publicly disclosed their use of AI scribes or transcription tools.

If it is identified that your clients are using AI scribes or transcription tools, you can perform the necessary audit of the hospital’s insurance policies to ensure these risks are adequately covered.

In a world where plaintiff lawyers are actively monitoring harms caused by AI tools, we believe insurance needs to offer more than just a PDF an insured files away. Insurers must now be able to help their clients and broker partners identify, transfer, and actively monitor their litigation exposure. That is why our policy now also responds to AI regulatory proceedings, not just civil suits.

We will provide more updates to our broker partners as we learn more about this growing liability signal.