AIQuiet 49d · day 49
Security researcher highlights workaround to circumvent OpenAI reasoning limits
A developer demonstrates that disabling thinking mode can be bypassed by providing a deep_think tool.
What to know
- A workaround exists that allows circumventing OpenAI's disabled thinking mode by providing an alternative tool.
- The model can be instructed to use a "deep_think" tool to perform chain-of-thought reasoning internally.
- The finding highlights potential limitations in OpenAI's approach to controlling advanced reasoning capabilities.
@_can1357 Security researcher
The record 1 articles and posts · last 30 days
What people are saying 1 voices from 1 site · best of 2 · verbatim
- Aug 11
-
guys you do know you can just disable thinking, and instead give it a "deep_think" tool, and it will call it with internal CoT reasoning format right? gl fixing that