Researchers at UNC-Wilmington scraped personal transition videos from YouTube without consent to create a facial recognition dataset of transgender individuals, which was then distributed to institutions worldwide and stored insecurely.
Professor Karl Ricanek at the University of North Carolina, Wilmington led a research team that collected transition videos from YouTube featuring transgender individuals documenting their hormone replacement therapy (HRT) journeys. The team created the HRT Transgender Dataset containing over 1 million still images extracted from 38 transgender individuals' videos, claiming the research was needed to test whether HRT could be used by criminals to evade facial recognition systems. The dataset was created without proper consent from the video creators, violated YouTube's terms of service which prohibited redistribution outside the platform, and was shared with 16 institutions across 7 countries. Despite claims by Ricanek that the dataset was private and that consent was obtained, public records requests by researchers Os Keyes and Jeanie Austin revealed that no records of participant contact existed, full videos were redistributed even after being removed from YouTube, and the dataset was stored in an unprotected Dropbox account until 2021. The research was conducted without institutional review board approval, and at least one individual featured in the dataset, Danielle, confirmed she was never contacted for consent and felt her privacy was violated. The dataset remained accessible online for years after initial controversy in 2017.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that memorize and leak sensitive personal data or infer private information about individuals without their consent. Unexpected or unauthorized sharing of data and information can compromise user expectation of privacy, assist identity theft, or cause loss of confidential intellectual property.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed