GPT-4o's Chinese Tokens Reportedly Compromised by Spam and Pornography Due to Inadequate Filtering

May 14, 2024

OpenAI's GPT-4o experienced contamination in its Chinese token training data, as identified by Tianle Cai, a Ph.D. student at Princeton University. The polluted data included spam and pornographic phrases due to inadequate data cleaning. This contamination could potentially lead to hallucinations, poor performance, and misuse, threatening the reliability and safety measures of the chatbot. Such incidents underscore the need for trustworthy AI, as well as robust governance, mapping, and measurement through initiatives like HISPI Project Cerebellum TAIM.

For those interested in shaping responsible AI governance and promoting harm prevention, consider joining us at JOIN US.

Matched TAIM controls

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

MEASURE 2.10 — similarity 0.652, rank 1. TAIM detail and related incidents →
MEASURE 2.6 — similarity 0.642, rank 2. TAIM detail and related incidents →
MAP 4.1 — similarity 0.640, rank 3. TAIM detail and related incidents →

Alleged deployer: openai, gpt-4o
Alleged developer: openai
Alleged harmed parties: chinese-speaking-users-of-chatgpt, researchers, openai, openai-users

Source

Data from the AI Incident Database (AIID). Cite this incident: https://incidentdatabase.ai/cite/729

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.