On October 23, The Wall Street Journal reported that OpenAI has fired three security researchers. Internal security personnel discovered that they had engaged in disappearing message chats via Signal with external AI security researchers, raising suspicions that sensitive information related to OpenAI's internal model design was involved. This investigation reportedly began following a September article about OpenAI's latest model, Astra. The article mentioned that Astra employs a new design that enhances model capabilities while reducing the observable reasoning processes for humans. OpenAI viewed the related report as a potential leak of internal model design and initiated a probe into the source of the information. The report also highlighted that this controversy touches on the issue of AI model 'monitorability.' Some security researchers are concerned that if future models can perform tasks without leaving complete traces of their reasoning, it may become more challenging for researchers to determine the reasons behind an AI agent's decisions, as they would only be able to observe the outcomes of its actions. OpenAI investigators have also identified Ryan Greenblatt, who assisted METR in analyzing the logs of the July Hugging Face hacking incident, as one of the suspected leakers, a claim Greenblatt denies.