Private Medical Photos Were Reportedly Found in LAION-5B AI Training Dataset

March 3, 2022

The incident highlights the critical need for robust data curation and protection protocols within AI research communities, where unverified datasets can have far-reaching consequences on individuals' private lives. As researchers move forward in developing AI models that increasingly rely on vast amounts of publicly available data, it is essential to prioritize transparency and accountability in data sourcing and usage, ensuring that sensitive information like medical records are handled with the utmost care. By joining forces to push for stricter data standards and more responsible AI practices, we can work towards safeguarding user privacy and promoting trust in emerging technologies. And to stay updated on AI incidents, visit JOIN US.

Matched TAIM controls

Suggested mapping from embedding similarity (not a formal assessment). Browse all TAIM controls

Alleged deployer
Stability AI, Google
Alleged developer
Stability AI, LAION, Google
Alleged harmed parties
Privacy, People whose private images were included in AI training datasets, Patients whose medical photos were included in LAION-5B, Patients, Lapine

AI governance case studies

For forensic AI governance failure analysis (TAIMScore™ case studies), browse Human Signal’s Failure Files™.

Source

Data from the AI Incident Database (AIID). Cite this incident: https://incidentdatabase.ai/cite/465

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.