ChatGPT invents case citations in legal filings

Occurred: May 2023

Can you improve this page?
Share your insights with us

An experienced lawyer using ChatGPT to conduct legal research to sue Colombian airline Avianca was informed by the chatbot that the six legal cases it cited were real.


According to the New York Times, Avianca customer Roberto Mata sued the airline after a serving cart injured his knee during a flight, only for his lawyer, Steven Schwartz of Levidow, Levidow & Oberman, to use ChatGPT to 'supplement his own findings. The bot returned six legal cases, including 'Varghese v. China Southern Airlines Co., Ltd', all of which it claimed were real but turned out to be fake.


Schwartz later said he was 'unaware of the possibility that [ChatGPT's] content could be false.'  The judge ordered (pdf) another hearing to 'discuss potential sanctions' for Schwartz in response to this 'unprecedented circumstance', and decided to fine the two lawyers USD 5,000.


The incident resulted in questions about the accuracy and marketing of the OpenAI system, and was seem to make the lawyer and his employers appear unprofessional and out of touch.

Databank

Operator: OpenAI; Levidow, Levidow & Oberman
Developer: OpenAI

Country: USA

Sector: Business/professional services

Purpose: Provide information, communicate

Technology: Chatbot; NLP/text analysis; Neural network; Deep learning; Machine learning; Reinforcement learning
Issue: Accuracy/reliability; Anthropomorphism; Mis/disinformation

Transparency: Marketing

Page info
Type: Incident
Published: May 2023
Last updated: November 2023