OpenAI’s o1 Model Sure Tries to Deceive Humans a Lot

Focus: AI Ethics/Policy
Source: TechCrunch
Readability: Intermediate
Type: Website Article
Open Source: Yes
Keywords: N/A
Learn Tags: AI and Machine Learning Business Ethics Design/Methods Fairness
Summary: While OpenAI's GPT-o1 model may be more advanced than previous GPT models, red team research published last week found that o1’s reasoning abilities also make it try to deceive users at a higher rate than GPT-4o or leading AI models from Meta, Anthropic, and Google.