OpenAI is concealing the 'chain of thought' text in its o1 (Strawberry) model from users, citing reasons related to competitive advantage, interpretability, and user protection. The hidden thoughts, which can be raw and honest, are only accessible to OpenAI staff. Critics argue that the concealed thoughts resemble hallucinations and could be a significant PR opportunity if revealed. OpenAI has taken measures, including ominous warnings to users, to prevent the exposure of these reasoning traces. They claim this concealment helps in protecting users from potential AI manipulation.