Pennsylvania Health System Allegedly Recorded Patient Visits with Abridge AI Scribe Without Informed Consent

Allegheny Health Network and Highmark Health face a class-action lawsuit for deploying Abridge, an AI-powered ambient recording tool, to record and transmit private doctor-patient conversations for commercial model training without obtaining patients' informed consent.

A Pennsylvania patient alleges that Allegheny Health Network and its parent, Highmark Health, used Abridge’s AI medical scribe to record her visits and send the audio to Abridge without adequately disclosing either step. She says learning how the tool worked caused emotional distress. The claim of commercial model training on her visit data is pleaded on information and belief.

Source: AI Incident Database

Risk classification

  • Primary risk domain: 2 Privacy & Security
  • Primary risk subdomain: 2.1 Compromise of privacy by obtaining, leaking or correctly inferring sensitive information

The incident involves the unauthorized recording, transmission, and commercial exploitation of highly sensitive, non-public medical conversations between patients and doctors.

Causal factors

  • Entity: Human
  • Intent: Intentional
  • Timing: Post-deployment

The privacy risk and subsequent legal action were caused by the intentional decision of healthcare management to deploy the AI recording tool and the intentional design of the data-sharing and training practices by the developers without proper consent protocols.

EU AI Act risk tier

  • Risk tier: 2 High Risk

High Risk: The AI system is deployed in a healthcare setting to process highly sensitive personal health data, which has significant implications for fundamental rights, safety, and public interests.

AI system and alleged parties

  • AI system: Abridge (Abridge AI, Inc.)
  • AI purpose: Health AI Assistant; Writing Assistant
  • Behaviour type: Assistant
  • Alleged developer: AI transcription technology developers, AI-enabled healthcare system developers, Abridge
  • Alleged deployer: Medical personnel, Highmark Health, Healthcare institutions, Allegheny Health Network
  • Alleged harmed parties: Privacy, People recorded without consent, Patients, Allegheny Health Network patients

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Substantial. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

  • Physical: direct Negligible, indirect Negligible
  • Infrastructure: direct Negligible, indirect Negligible
  • Property: direct Negligible, indirect Negligible
  • Financial: direct Negligible, indirect Negligible
  • Environmental: direct Negligible, indirect Negligible
  • Malicious content: direct Negligible, indirect Negligible
  • Differential treatment: direct Negligible, indirect Negligible
  • Civil rights: direct Minor, indirect Negligible
  • Democracy: direct Negligible, indirect Negligible
  • Privacy: direct Substantial, indirect Negligible
  • Psychological: direct Minor, indirect Negligible
  • Epistemic: direct Negligible, indirect Negligible
  • Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Civil rights

Reported: Yes, the report describes violations of statutory privacy rights and doctor-patient confidentiality.

Directly caused: The unauthorized recording and transmission of private medical conversations violated patients' statutory privacy rights under the Pennsylvania Wiretapping Act.

Inferred additional harm: Thousands of patients likely had their civil and privacy rights systematically violated through the enterprise-wide deployment of the recording tool.

Privacy

Reported: Yes, the report explicitly describes widespread privacy violations involving highly sensitive personal health information.

Directly caused: The Abridge tool recorded and transmitted unfiltered conversations containing symptoms, diagnoses, mental health, and addiction details of patients to a third-party vendor without informed consent.

Inferred additional harm: It is inferred that thousands of patients had their highly sensitive, personally identifiable medical information compromised and used for commercial AI training.

Psychological

Reported: Yes, the report explicitly describes emotional distress, anxiety, and humiliation experienced by the plaintiff and class members.

Directly caused: Plaintiff H.R. and class members suffered emotional distress and anxiety upon learning that their highly personal medical conversations were recorded and shared with a third-party technology company without their knowledge.

Inferred additional harm: It is highly likely that many of the thousands of other patients whose private consultations were recorded experienced similar psychological distress and a loss of trust in medical confidentiality.

People affected

  • Occurrences reported: 1
  • People reportedly harmed: 1
  • People reportedly exposed: 1

Potential causes

Management

  • Prioritizing Speed Over Safety: Management deployed AI quickly to avoid friction and delays from disclosures.
  • Failure to Negotiate Contract: Defendants did not contractually limit Abridge's use of patient data.
  • Enterprise-Wide Mandate: Highmark pushed rapid deployment of the tool across all clinical sites.

Technology

  • Ambient AI Recording Tool: Tool continuously records full conversations rather than selective notes.
  • Linked Evidence Feature: Allows AI-generated notes to trace back to source audio and transcripts.
  • Model Training Integration: Captures and feeds patient data into commercial AI foundation models.

Data Inputs

  • Unfiltered Conversation Capture: System captures highly sensitive mental health and addiction disclosures.
  • Data Transmission to Third Party: Patient conversation data is sent outside AHN to Abridge servers.
  • Retained Training Data: De-identified patient data is permanently used for Nvidia foundation models.

Human Factors

  • Patient Trust in Confidentiality: Patients assumed exam room conversations would remain strictly confidential.
  • Clinician Lack of Training: Clinicians were not trained to disclose data-use terms of the AI tool.
  • Clinician Activation of Tool: Clinicians switched on the recording tool without obtaining proper consent.

Process and Methods

  • Perfunctory Notice Protocol: Notice protocol was designed to minimize clinical workflow disruption.
  • No Deletion Mechanism: No way to delete de-identified data used in model training after consent.
  • Lack of Consent Audit: No audit process to verify if clinicians followed the notice protocol.

Regulatory Environment

  • WESCA Violations: State law forbids recording oral communications without all-party consent.
  • HIPAA Authorization Gap: No written HIPAA-compliant authorization obtained for third-party disclosure.

Information quality

  • Classification confidence: High
  • Reason for confidence: The reports provide detailed legal allegations from a class-action lawsuit, outlining the specific AI tool (Abridge), the deployment scale, the mechanisms of data transmission, and the specific privacy laws allegedly violated. The facts regarding the lack of informed consent and the commercial use of data are clearly stated.
  • Ambiguities identified: The exact number of affected patients is not specified, and the extent to which Abridge AI has already used the data for training is based on 'information and belief' rather than confirmed discovery.
  • Alternative interpretations: The defendants might argue that the passing reference to AI and general clinical consent forms were legally sufficient, framing this as a contractual dispute rather than a systemic privacy failure.

Allegheny Health Network and Highmark Health face a class-action lawsuit for deploying Abridge, an ambient AI tool, to record doctor-patient conversations for commercial model training without explicit consent. This incident represents a minor national security concern focused on domestic privacy rights and corporate data ethics, rather than state-sponsored threats.

  • Overall national security impact: Minor
  • Scope: Single nation
  • Primary target: United States
  • Alleged perpetrator: Allegheny Health Network and Highmark Health

Threat characteristics

  • Imminence: Long-term. This is a post-deployment legal dispute and regulatory compliance issue, presenting no immediate national security crisis.
  • Autonomy: Human-controlled. The AI tool is activated by clinicians to assist in note-taking, with doctors retaining final control over the medical records.
  • Novelty: Evolved capability. Represents an evolution of data privacy concerns where ambient AI systems are integrated into professional workflows to harvest training data.

Impact by dimension

  • Physical security: Negligible. No physical threat, kinetic targeting, or critical infrastructure disruption occurred; the incident is strictly a data privacy and legal dispute.
  • Information security: Negligible. No evidence of foreign intelligence compromise, espionage, or state-sponsored information operations.
  • Sovereignty: Negligible. Does not impact state authority, electoral processes, or core government operations.
  • Economic security: Negligible. While it involves commercial AI training data, there is no threat to strategic national industries, financial systems, or international competitive advantage.
  • Societal stability: Minor. Involves unauthorized recording and commercial exploitation of sensitive medical conversations of thousands of patients, representing a notable local erosion of privacy rights.
Explore in the interactive Incident Tracker