OpenAI has recently terminated the employment of three security researchers. This move came after internal security staff found that these researchers had used Signal's self-destructing message function to communicate with external AI safety researchers. This raised suspicions that they might have leaked sensitive information, such as the internal model designs of OpenAI. The investigation was prompted by a September report that discussed OpenAI's new model, Astra. This model featured a novel design aimed at boosting its capabilities while minimizing the observable reasoning processes for humans. OpenAI suspected that the report might have included leaked details about its internal model designs, prompting the launch of an investigation. Moreover, the controversy also touches upon the monitorability of AI models. Some security researchers worry that if future models can execute tasks without leaving full reasoning trails, it will become increasingly challenging for researchers to understand the rationale behind AI Agents' decisions. Among those suspected of leaking information was Ryan Greenblatt, who had previously helped METR analyze logs from the July Hugging Face hack. However, he refuted these allegations.