X Fails to Remove AI-Generated Deepfake Nudes Reported as Nonconsensual Content
A preprint study by University of Michigan and Florida International University researchers found that X removes AI-generated deepfake nude images when reported as copyright violations but fails to remove them when reported under its nonconsensual nudity policy, suggesting the platform's moderation systems treat copyright complaints with greater urgency.
What happened
Researchers at the University of Michigan and Florida International University conducted an audit of X's content removal systems using AI-generated deepfake nude images. They created 50 images of five AI personas of young white women and posted them across 10 accounts, then filed reports on 25 images as DMCA copyright violations and 25 as nonconsensual nudity.
All 25 images reported as copyright violations were removed within 25 hours, and the accounts that posted them received temporary suspensions. None of the 25 images reported as nonconsensual nudity were removed after a three-week waiting period, and the accounts posting them faced no consequences.
Reported impact
The study revealed a significant disparity in X's effectiveness between DMCA-based takedown requests and reports filed under the platform's internal nonconsensual nudity policy.
Index analysis
The findings demonstrate that X's moderation systems prioritize copyright enforcement over protection against nonconsensual sexual imagery, despite both violating the platform's policies.
Governance controls
X owner Elon Musk dissolved the platform's trust and safety council in 2022. The site has since reopened two dozen safety and cybersecurity positions in the U.S.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.