Cigna Algorithm PXDX Allegedly Rejected Thousands of Patient Claims En Masse in Breach of California Law

Cigna health insurer faces a class-action lawsuit for allegedly using the PXDX ("procedure-to-diagnosis") algorithm to automatically reject over 300,000 patient claims in violation of California law, prompting two members to file the lawsuit seeking damages and a jury trial. Cigna disputes the allegations, claiming the process expedites physician reimbursement and does not result in care denials.

Risk classification

• Primary risk domain: 5 Human-Computer Interaction
• Primary risk subdomain: 5.1 Overreliance and unsafe use

Cigna's doctors relied entirely on the PXDX algorithm's automated flags to reject claims in bulk, spending only 1.2 seconds per claim without opening patient files.

Additional risk subdomains

• 7.3 Lack of capability or robustness: The algorithm systematically flagged and rejected valid medical claims, failing to perform reliable and fair individual assessments.

Causal factors

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

The automated claims denials were caused by the active operation of Cigna's deployed PXDX algorithmic system.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The system is used in healthcare and insurance decision-making, which has significant implications for safety, fundamental rights, and access to essential services.

AI system and alleged parties

• AI system: PXDX
• AI purpose: Application Evaluation; Financial Processing
• Behaviour type: Assistant
• Alleged developer: Cigna
• Alleged deployer: Cigna
• Alleged harmed parties: 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 Minor, 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 Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Financial

Reported: The report explicitly describes financial losses, noting that patients were forced to pay for tests out-of-pocket after Cigna denied their claims.

Directly caused: The average financial loss per occurrence is not specified, but individual patients were forced to pay out-of-pocket for denied medical tests.

Inferred additional harm: Across more than 300,000 denied claims, the total financial loss to patients is likely in the millions of dollars.

Civil rights

Reported: The report explicitly describes violations of rights, alleging that Cigna violated California requirements for thorough, fair, and objective investigations of medical bills.

Directly caused: The automated system denied patients their legal right to an individual, fair review of their insurance claims.

Inferred additional harm: Thousands of patients had their consumer rights systematically violated by the automated denial process.

People affected

• Occurrences reported: 1
• People reportedly harmed: 2000
• People reportedly exposed: 300000

Potential causes

Management

• Prioritizing Speed Over Care: Management prioritized rapid, automated processing to reduce operational costs.
• Exploiting Low Appeal Rates: Management knew only a small fraction of policyholders appeal denied claims.

Technology

• Algorithmic Matching Limitations: PXDX only flags code discrepancies without understanding medical necessity.
• Lack of Medical File Integration: The software does not automatically retrieve or analyze full patient records.

Data Inputs

• Reliance on Billing Codes Only: System relies solely on codes, ignoring clinical context.

Human Factors

• Automation Bias by Doctors: Doctors signed off on denials without opening patient medical files.
• Negligible Human Review Time: Doctors spent an average of 1.2 seconds reviewing each flagged claim.

Process and Methods

• Batch Claim Processing: Claims were processed in massive batches, preventing individual evaluation.
• Inadequate Review Workflow: Workflow did not require doctors to open or read patient medical records.

Regulatory Environment

• Unenforced Insurance Codes: Lack of immediate enforcement of state laws requiring thorough investigations.

Information quality

• Classification confidence: High
• Reason for confidence: The reports from Forbes and AP provide clear, consistent details about the lawsuit, the PXDX system's function, and the scale of the denials. Cigna's counter-arguments are also clearly presented, allowing for a balanced assessment.
• Ambiguities identified: There is a conflict between the plaintiffs' claim that the system automatically denies necessary care and Cigna's claim that it is a post-treatment administrative tool that does not result in denials of care.
• Alternative interpretations: The system could be viewed purely as an administrative database tool rather than an AI safety risk, but the automated decision-making aspect and its impact on patient health support the AI safety classification.

The incident involves a class-action lawsuit against Cigna for using its PXDX algorithm to automatically deny over 300,000 medical claims with minimal human review. While presenting significant consumer rights and healthcare access concerns, the incident remains a civil regulatory matter with minor national security implications.

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

Threat characteristics

• Imminence: Long-term. Represents an ongoing regulatory and legal concern regarding corporate AI deployment rather than an immediate crisis.
• Autonomy: Human-supervised. The system flagged claims for denial, which were then rapidly signed off by human doctors, representing human-supervised automation with minimal actual review.
• Novelty: Evolved capability. Demonstrates a significant scaling of automated decision-making to process and deny medical claims, representing an evolution of corporate algorithmic use.

Impact by dimension

• Physical security: Negligible. Incident involves private health insurance claims processing, not threats to physical critical infrastructure, kinetic systems, or essential service supply chains.
• Information security: Negligible. No elements of information warfare, intelligence compromise, or systematic disinformation are present in this incident.
• Sovereignty: Negligible. Represents a civil legal dispute involving a private corporation and state regulations, with no threat to state authority or core government operations.
• Economic security: Negligible. While causing individual financial burdens, the incident does not threaten national economic stability, strategic industries, or technological competitive advantage.
• Societal stability: Minor. Widespread automated denial of medical claims impacted thousands of citizens' consumer rights, but the societal impact is limited and managed through standard civil legal procedures.