OpenAI's GPT-2 Memorizes PII in Training Data: A Case for Safe and Secure AI

February 14, 2019

The incident involving OpenAI's GPT-2 highlights the importance of trustworthy AI. The model memorized and could recite verbatim instances of training data, potentially exposing personally identifiable information (PII) such as names, emails, twitter handles, and phone numbers. This AI incident maps to the Govern function in HISPI Project Cerebellum Trusted AI Model (TAIM). Join us in upholding responsible AI governance by contributing to our AI incident database and helping shape a safer digital future. JOIN US
Alleged deployer
openai
Alleged developer
openai
Alleged harmed parties
openai, people-having-personal-data-in-gpt-2's-training-data

Source

Data from the AI Incident Database (AIID). Cite this incident: https://incidentdatabase.ai/cite/357

Data source

Incident data is from the AI Incident Database (AIID).

When citing the database as a whole, please use:

McGregor, S. (2021) Preventing Repeated Real World AI Failures by Cataloging Incidents: The AI Incident Database. In Proceedings of the Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (IAAI-21). Virtual Conference.

Pre-print on arXiv · Database snapshots & citation guide

We use weekly snapshots of the AIID for stable reference. For the official suggested citation of a specific incident, use the “Cite this incident” link on each incident page.