What happens if AI agents start speaking their own language? | CNN Business

CNN Business published an article headlined "What happens if AI agents start speaking their own language?" According to the article, OpenAI said in a statement last week that it is investing "significant resources" in the ability to monitor AI models. The article reports that the company fired three safety and alignment researchers, and that OpenAI said the dismissals were not tied to employees voicing safety concerns.

The article also describes the work of researchers at METR, who pore over AI agents' "chain of thought" scratch pads to see the steps bots took before responding to a request. According to the article, that work was key to its investigation of the Hugging Face hack this year. The piece presents this account of how model reasoning is currently inspected alongside OpenAI's statement about monitoring resources.

The article's title poses the question of what happens if AI agents begin speaking a language of their own. The reported material it presents concerns OpenAI's stated investment in monitoring capability following the dismissals, and METR's examination of agent chain-of-thought reasoning as applied in the Hugging Face hack investigation this year.