Perplexity AI Reportedly Misstated CLL Research, Allegedly Contributing to Delayed Treatment and Prolonged Suffering

Joseph Neal Riley reportedly used Perplexity AI to self-diagnose a rare complication of chronic lymphocytic leukemia (CLL) and, based on that output, allegedly delayed an oncologist-recommended Ven-Obi treatment for about a year. His son, Benjamin Riley, alleges that Perplexity misstated the medical research it cited, a conclusion he purportedly later confirmed with the study's authors.

Risk classification

• Primary risk domain: 3 Misinformation
• Primary risk subdomain: 3.1 False or misleading information

The AI system generated false and misleading medical information that led the patient to make an unsafe decision regarding his cancer treatment.

Additional risk subdomains

• 5.1 Overreliance and unsafe use: The patient relied on the AI system for critical medical decisions, bypassing professional medical advice.
• 7.3 Lack of capability or robustness: The AI system failed to accurately summarize the medical research paper, demonstrating a lack of capability in factual retrieval.

Causal factors

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

The incident was caused by the Perplexity AI system generating incorrect medical information after its deployment.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The system is a chatbot/conversational search engine, which falls under transparency obligations for interactive AI systems.

AI system and alleged parties

• AI system: None named
• AI purpose: Question Answering; Health AI Assistant
• Behaviour type: Assistant
• Alleged developer: Perplexity AI
• Alleged deployer: Perplexity AI
• Alleged harmed parties: People seeking medical advice, Joseph Neal Riley, Epistemic integrity, Benjamin Riley

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Minor. 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 Substantial
• 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 Negligible, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Minor
• Epistemic: direct Minor, indirect Minor
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Physical

Reported: The report explicitly describes physical harm to the author's father caused indirectly by the incident.

Indirectly caused: The father delayed his Ven-Obi cancer treatment for a year based on the AI's incorrect advice, which likely hastened his declining health, magnified his pain, and culminated in his death.

Psychological

Reported: The report explicitly describes psychological distress and frustration caused indirectly by the incident.

Indirectly caused: The author experienced significant distress and frustration upon discovering the father's reliance on the AI. The father suffered from anxiety and the false belief of having a rare, painful complication.

Epistemic

Reported: The report explicitly describes epistemic harm caused directly by the incident.

Directly caused: Perplexity AI generated a report that misstated the conclusion of medical research, falsely claiming that refraining from Ven-Obi treatment was proper for Richter's Transformation.

Indirectly caused: The father believed this false information, leading him to reject recommended medical treatment.

People affected

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

Potential causes

Management

• Prioritizing Utility Over Safety: Promoting AI as a health resource despite known factual unreliability.

Technology

• AI Hallucination of Facts: The model generated non-factual statements and misstated research findings.
• Authoritative Tone of Output: The AI strikes an authoritative tone that makes errors seem highly credible.

Data Inputs

• Misrepresented Research Data: The AI incorrectly summarized the conclusion of the cited medical study.

Human Factors

• Automation Bias in Self-Diagnosis: The patient trusted the AI's diagnosis over his oncologist's recommendation.
• Confirmation Bias Amplification: The AI's output affirmed the patient's existing fear of hospital treatments.

Process and Methods

• Inadequate Output Verification: The system relies on users to verify complex and critical medical outputs.

Regulatory Environment

• Lack of Medical AI Regulation: Absence of strict enforcement preventing AI from giving unlicensed advice.

Information quality

• Classification confidence: High
• Reason for confidence: The report provides a detailed, first-hand account of the incident, specifying the AI system used (Perplexity AI), the exact medical context (CLL and Ven-Obi treatment), and the specific error made by the AI (misstating the conclusion of a medical study). The chain of causality from the AI's output to the patient's decision and subsequent health decline is clearly documented.
• Ambiguities identified: The exact date of the incident is not specified, and the author notes it is impossible to prove with absolute certainty whether the father would have rejected treatment even without the AI's output.
• Alternative interpretations: The incident could be viewed primarily as a case of human medical non-compliance driven by pre-existing fear of hospitals, with the AI merely serving as a convenient justification rather than the primary cause.

A private citizen tragically passed away after delaying cancer treatment based on incorrect medical research summaries generated by Perplexity AI. While representing a serious consumer safety and AI reliability issue, the incident has negligible implications for national security.

• Overall national security impact: Negligible
• Scope: Single nation
• Primary target: No clear primary
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. Represents an ongoing, long-term challenge regarding the reliability of AI-generated medical information rather than an imminent national security crisis.
• Autonomy: Human-controlled. The AI acted as an information retrieval assistant, with the human user retaining full control over his own medical decisions.
• Novelty: Established threat. AI hallucination and the propagation of incorrect medical or factual information is a well-documented and established threat.

Impact by dimension

• Physical security: Negligible. The physical harm was confined to a single individual's tragic medical decision and has no systemic impact on national physical security or critical infrastructure.
• Information security: Negligible. No evidence of state-sponsored disinformation, intelligence compromise, or systematic information warfare campaign.
• Sovereignty: Negligible. The incident involves a private citizen's healthcare decisions and does not impact government functions, elections, or state sovereignty.
• Economic security: Negligible. While reflecting a technical failure in LLM summarization, this isolated incident does not threaten national economic stability or strategic technological competitive advantage.
• Societal stability: Negligible. This is an isolated, individual tragedy rather than a systematic threat to civil liberties, population-scale human rights, or societal stability.