LLMMedium
GPT-4o's Chinese Tokens Reportedly Compromised by Spam and Pornography Due to Inadequate Filtering
- Occurred
- May 14, 2024
- Published
- May 17, 2024
- Source
- AI Incident Database
Summary
OpenAI's GPT-4o was found to have its Chinese token training data compromised by spam and pornographic phrases due to inadequate data cleaning. Tianle Cai, a Ph.D. student at Princeton University, identified that most of the longest Chinese tokens were irrelevant and inappropriate, primarily originating from spam and pornography websites. The polluted tokens could lead to hallucinations, poor performance, and potential misuse, undermining the chatbot's reliability and safety measures.
Developers
- OpenAI
Deployers
- OpenAI
- GPT-4o
Harmed parties
- Researchers
- OpenAI
- OpenAI users
- ChatGPT
Classification
Category confidence: 95%
Severity confidence: 90%
Rationale: GPT-4o training data contamination affects language model reliability and safety without immediate physical harm.
Classifier: Qwen/Qwen3.5-35B-A3B-FP8