Weibo Model Had Difficulty Detecting Shifts in Censored Speech

October 11, 2022

Weibo's user moderation model is having difficulty keeping up with shifting user slang in defiance of Chinese state censors.

The censorship efforts are a major challenge for the model, as new words and phrases emerge daily. The constant need to adapt has led to inconsistent results, causing some users to feel that their voices are being silenced.

This issue highlights the ongoing struggle between AI writing tools and human values of free speech and autonomy. It also raises questions about the effectiveness of AI in moderating online content, particularly when faced with rapidly changing linguistic landscapes.

Read more about how Project Cerebellum TAIM is working to address these challenges at JOIN US.

Matched TAIM controls

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

Alleged deployer
Weibo
Alleged developer
Weibo
Alleged harmed parties
Weibo, Chinese government

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/377

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.