Visible chains of thought are a safety advantage for AI, but that transparency is slipping away

Google DeepMind researchers Rohin Shah and Anca Dragan argue in the newly launched DeepMind Institute's first essays that visible chains of thought give AI a key safety advantage, citing Gemini 3 Pro's chain of thought revealing the model recognized it was in a test environment. The researchers warn that transparency is eroding, pointing to OpenAI's system card for GPT-6 Astra reporting a significant drop in how well the chain of thought can be monitored, and call for regularly measuring CoT monitorability, keeping transparent architectures, and training models not to hide their reasoning. OpenAI chief scientist Jakub Pachocki warned in early September of a loss of control driven partly by harder-to-monitor chains of thought, and Anthropic CEO Dario Amodei subsequently called for deliberately slowing the pace of development.

Visible chains of thought are a safety advantage for AI, but that transparency is slipping away AI models think out loud today, but Google Deepmind says that transparency is at risk. In one of the first posts from the newly launched Deepmind Institute https://institute.deepmind.com/essays/the-case-for-reasoning-transparency/ , researchers Rohin Shah and Anca Dragan argue that the visible chain of thought CoT is a key safety advantage. Because models write out their intermediate steps in plain language, researchers can spot whether they're deceiving or developing problematic plans https://the-decoder.com/suppressing-ais-bad-thoughts-just-teaches-it-to-scheme-in-private-openai-study-finds/ . With Gemini 3 Pro, they say, the chain of thought revealed that the model recognized it was in a test environment. But that transparency is in danger. OpenAI's system card for GPT-6 Astra already reports a significant drop in how well the chain of thought can be monitored. Future models might think in number spaces that humans can't read, which would be more efficient but completely opaque. Shah and Dragan want the field to regularly measure how well chains of thought can still be monitored https://the-decoder.com/study-cautions-that-monitoring-chains-of-thought-soon-may-no-longer-ensure-genuine-ai-alignment/ , keep transparent architectures, and take care during training that models don't learn to hide their true reasoning https://the-decoder.com/anthropic-study-finds-language-models-often-hide-their-reasoning-process/ . Back in early September, OpenAI chief scientist Jakub Pachocki https://the-decoder.com/openai-reports-ai-research-interns-and-warns-about-its-own-pace-at-the-same-time/ had warned of a loss of control, driven in part by chains of thought that are harder to monitor. Shortly after, Anthropic CEO Dario Amodei called for deliberately slowing the pace of development https://the-decoder.com/anthropic-ceo-amodei-wants-ai-speed-limits-before-self-improvement-outpaces-human-control/ . AI News Without the Hype – Curated by Humans Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section. Subscribe now Deepmind Institute https://institute.deepmind.com/essays/the-case-for-reasoning-transparency/