When you buy through affiliate links in our content, we may earn a commission at no extra cost to you. Learn 
OpenAI has taken a firm stance against users attempting to explore the inner workings of its newly launched AI model family, Strawberry.
The o1-preview and o1-mini models, which are said to possess advanced reasoning abilities, have sparked intense interest within the AI community. However, attempts to uncover the raw chain of thought driving these models have been met with strong warnings from OpenAI.
OpenAI’s Secretive Approach to AI Reasoning
Unlike its predecessors, such as GPT-4o, the o1 model employs a step-by-step problem-solving process that’s partially visible to users. The platform allows users to see a simplified version of this process but keeps hidden the raw chain of thought.
OpenAI claims that these hidden reasoning chains are essential for internal monitoring and maintaining a competitive edge.
This secrecy has led to a wave of jailbreak attempts, with AI enthusiasts and red-team researchers seeking to expose the underlying logic.
While early reports suggest minor breakthroughs, OpenAI remains vigilant. Users attempting to query the model’s reasoning are reportedly receiving warning emails, including from OpenAI’s internal safety teams.
Several users, including prominent AI prompt engineers like Riley Goodside, have reported receiving warnings for using terms like “reasoning trace” during interactions with the model.
Can confirm https://t.co/dfVGCJyCrK pic.twitter.com/KSw0Exbul8
— Lukas (@LukasBogacz) September 13, 2024
OpenAI’s warnings emphasize that such attempts violate the company’s Terms of Use and may result in a ban. The company advises users to stop engaging in activities designed to bypass its safeguards.
Marco Figueroa, Mozilla’s GenAI bug bounty program manager, shared his frustration on X, revealing that he was warned after his attempts to red-team the model. Figueroa expressed concerns that these restrictions hinder his ability to conduct safety research.
I was too lost focusing on #AIRedTeaming to realized that I received this email from @OpenAI yesterday after all my jailbreaks! #openAI we are researching for good!
You do have a safe harbor on your site https://t.co/R2UChZc9RO
and you have a policy implemented with… pic.twitter.com/ginDvNlN6M— MarcoFigueroa (@MarcoFigueroa) September 13, 2024
The Business Behind Hiding AI Reasoning
OpenAI’s decision to keep the raw reasoning hidden is seen by many as a strategic move.
The company acknowledges the potential disadvantages but claims that exposing the chain of thought could allow competitors to train similar models using OpenAI’s proprietary reasoning data.
Independent AI researcher Simon Willison echoed this sentiment, highlighting the potential competitive risks for OpenAI.
As OpenAI navigates growing commercial pressures and safety concerns, the company remains committed to controlling how its AI models are probed, ensuring its innovations remain protected from competitors.
