Google DeepMind stated that transparency in AI reasoning is under threat, according to The Decoder.
According to the brief description of the article, modern AI models currently demonstrate their chains of thought, but this openness is gradually weakening.
The significance of this trend cannot yet be assessed in detail: the available source does not disclose specific models, reasons for the changes, or possible countermeasures.
editorial commentary
Why it matters
A likely consequence is the complication of external verification of model behavior if visible chains of thought are reduced. The next observable signal will be specific examples of models, practices, or recommendations provided in the full article. Significant uncertainty remains: only the metadata synopsis from The Decoder is available to the editors.