California Health System Allegedly Used an AI Medical Scribe to Record Patients Without Consent

A proposed class-action lawsuit alleged that Sharp HealthCare, a California health system, used Abridge, an AI medical scribe, to record patient-clinician conversations without notice or consent and send sensitive audio to the vendor. The documentation workflow allegedly inserted false statements that patients had been informed and consented. Patient Jose Saucedo said he discovered the recording through his medical notes.

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 AI system recorded highly confidential doctor-patient conversations containing sensitive medical details and transmitted them to external servers without patient authorization.

Additional risk subdomains

• 3.1 False or misleading information: The AI tool automatically generated and inserted false statements into medical charts claiming patients had consented to the recording.

Causal factors

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

The primary privacy risk was caused by the human decision of Sharp HealthCare to intentionally deploy the AI recording tool without establishing a proper patient consent process.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The AI system is used in healthcare to document clinical interactions and generate medical notes, which directly impacts patient health records and falls under high-risk AI applications.

AI system and alleged parties

• AI system: Abridge
• AI purpose: Health AI Assistant; Technical Text Generation
• Behaviour type: Assistant
• Alleged developer: AI transcription technology developers, AI-enabled healthcare system developers, Abridge
• Alleged deployer: Sharp Rees-Stealy Medical Group, Sharp HealthCare, Medical personnel, Healthcare institutions
• Alleged harmed parties: Sharp HealthCare patients, Privacy, People recorded without consent, Patients, Jose A. Saucedo, Epistemic integrity

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 Substantial, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Substantial, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Minor, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Civil rights

Reported: Yes, the report describes violations of California privacy laws and the Confidentiality of Medical Information Act.

Directly caused: The unauthorized recording of confidential medical consultations directly violated patients' legal privacy rights.

Inferred additional harm: Up to 100,000 patients may have had their civil and privacy rights violated during the rollout of the AI tool.

Privacy

Reported: Yes, the report explicitly describes unauthorized recording and sharing of confidential medical conversations.

Directly caused: The AI app recorded exam room conversations containing symptoms, diagnoses, and personal details, and sent the audio to the vendor's servers without patient consent.

Inferred additional harm: The private medical data of over 100,000 patients was exposed to third-party employees on the vendor's servers.

Epistemic

Reported: Yes, the report describes the AI system fabricating consent statements in medical records.

Directly caused: The AI tool automatically inserted false statements into medical charts claiming patients were advised of and consented to the recording.

Inferred additional harm: The fabrication of consent records compromises the integrity and truthfulness of official medical documentation for thousands of patients.

People affected

• Occurrences reported: 1
• People reportedly harmed: 1
• People reportedly exposed: 100000

Potential causes

Management

• Hasty AI Tool Rollout: Sharp deployed Abridge without establishing proper consent protocols.
• Poor Vendor Oversight: Management failed to audit the AI's automated note-drafting behaviors.

Technology

• Automated False Documentation: AI automatically inserted incorrect consent statements into patient charts.
• Unsecured Cloud Transmission: Audio sent to vendor servers where employees could review sensitive data.

Data Inputs

• Missing Consent Verification Data: System lacked structured data inputs to verify actual patient consent.

Human Factors

• Failure to Inform Patients: Clinicians did not notify patients or seek consent before using the AI.

Process and Methods

• Flawed Note Automation Process: AI note-generation process assumed consent and hardcoded it into drafts.
• Lack of Opt-Out Mechanism: No process was established for patients to opt-out of AI recording.

Regulatory Environment

• Strict Two-Party Consent Laws: California law mandates consent from all parties before recording.
• Medical Privacy Regulations: CMIA requires written authorization before sharing patient data with vendors.

Information quality

• Classification confidence: High
• Reason for confidence: The report clearly details the parties involved, the specific AI tool (Abridge), the nature of the privacy violation, and the specific technical failure of fabricating consent notes.
• Ambiguities identified: It is not fully clear whether the fabrication of consent notes was a deliberate design choice by the developers or an algorithmic error/hallucination.

A class-action lawsuit alleges Sharp HealthCare deployed the Abridge AI app to record doctor-patient conversations without consent and automatically fabricate consent statements in medical records. While raising significant domestic privacy and data integrity issues for over 100,000 patients, the national security impact remains negligible.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: United States
• Alleged perpetrator: Sharp HealthCare

Threat characteristics

• Imminence: Long-term. The incident represents a localized legal and regulatory compliance issue rather than an active national security crisis.
• Autonomy: Human-supervised. The AI system operates as an assistant to draft notes, but executed the recording and consent fabrication steps autonomously within its workflow.
• Novelty: Evolved capability. Demonstrates an evolved risk where generative AI assistants fabricate legal or administrative compliance data within official records.

Impact by dimension

• Physical security: Negligible. No kinetic threats, physical system compromises, or critical infrastructure disruptions were reported in connection with this healthcare AI deployment.
• Information security: Negligible. No compromise of national intelligence, state-sponsored information operations, or classified systems occurred.
• Sovereignty: Negligible. The incident is a domestic civil law matter involving a private healthcare provider and does not impact state sovereignty or government operations.
• Economic security: Negligible. While there are legal and financial risks for the involved entities, there is no threat to national economic stability or strategic technological advantage.
• Societal stability: Minor. The unauthorized recording of up to 100,000 patients and AI-facilitated fabrication of consent records represents a notable domestic privacy and civil rights concern, though manageable via standard legal channels.