Source: some of the information that the three OpenAI employees allegedly mishandled pertained to OpenAI's infrastructure architecture (Rachel Metz/Bloomberg)

OpenAI terminated three researchers after they disclosed confidential infrastructure architecture details to an external AI safety organization, violating the company’s handling-sensitive-information policy. The employees had shared data that included proprietary system designs and internal security protocols. The dismissal follows a review that found the disclosures breached OpenAI’s confidentiality agreements.