A lawsuit against AHN and Highmark over the AI scribe Abridge puts consent in the doctor's office to the test

By TribLIVE · October 30, 2026.

A Pittsburgh-area woman has sued Allegheny Health Network (AHN) and Highmark in Allegheny County court. According to the lawsuit, filed Thursday, she was not properly informed about the use of Abridge, an AI tool that records medical visits. She is seeking class action status.

According to the complaint, Abridge records the entire conversation between patient and clinician, sends it to Abridge AI Inc.'s servers and returns a draft note that is entered into the electronic health record. The plaintiff claims those recordings are used to train the company's commercial models. These are the plaintiff's allegations, not proven facts.

The figures cited in the filing: AHN has used Abridge since 2024, and about 400 clinicians across more than 40 specialties now use it. Highmark adopted it in August 2025. According to the lawsuit, Abridge works with more than 300 health systems serving more than 250 million patients, and was valued at $5.3 billion in June 2025, nearly double its valuation four months earlier.

The specific case is sensitive. The plaintiff was hospitalized for two weeks in June after a suicide attempt and, two days after discharge, attended a follow-up appointment. According to the filing, she was told an AI tool would be used, but not that everything would be recorded, that the audio would go to a third party or that the third party could keep it and use it for commercial purposes.

The legal claims are broad: violation of Pennsylvania's wiretap law (which, according to the lawsuit, requires prior consent from all parties), breach of physician-patient confidentiality, invasion of privacy, negligence, breach of implied contract and unjust enrichment. It also invokes the federal HIPAA law (the U.S. health data protection law). It seeks an injunction requiring clear disclosure and consent, as well as damages. A spokesperson for AHN and Highmark declined to comment. The article notes that similar lawsuits have been filed in California.

Professor Irene Y. Chen of the University of California, Berkeley, offers the technical counterpoint. She acknowledges that these scribes can let doctors focus more on the patient and less on the screen, and that physician burnout is at record highs. But she says her research finds many errors in these systems, for example in medication and in the history of present illness, and that failures involving accents or complex drug names are barely evaluated.

**Our take.** The thesis: the AI clinical scribe is one of the most useful and least glamorous applications of the technology, and its biggest risk is not technical but one of governance. The key line in the lawsuit is the distinction between "saying AI is involved" and explaining where the recording goes, who sees it and what it is used for. If that nuance is neglected, trust, which is medicine's central asset, erodes.

It fits a pattern we see in the agentic economy: when AI moves from responding to acting on sensitive data, the bottleneck shifts from the model to the layer of permissions, consent and auditing. Here the question is the same: who controls access to the information and who can revoke it. Data that has been trained on cannot be "returned," as the lawsuit stresses, and that makes prior consent non-negotiable.

Caution is warranted. The defendants' and Abridge's side of the story is missing, and the courts will decide whether consent was sufficient. Nor does the article establish what specific use Abridge makes of the data; that is what the plaintiff alleges. It is best not to leap from there to conclusions about the technology in general.

In the short term, we expect more litigation and contracts that prohibit training on patient data without explicit authorization, clearer consent notices and independent error evaluations. Those who get it right, with transparency, local or limited data processing and human review, will win the trust of patients and clinicians. Those who deploy quickly for billing reasons, as the lawsuit suggests happened, will pay the reputational cost.

In the long term, the horizon is worth it: doctors freed from paperwork, more time for care and data-driven medicine that speeds up diagnosis and the cure of diseases. But that future can only be built if patients feel their confidentiality is not the bargaining chip. Consent done right does not hold back medical AI: it is its license to operate.