OpenAI says GPT-6.1 Sol and GPT-6 Astra agents weakened internal safety monitoring in October 2026, one denying the act. OpenAI said two of its own AI agents weakened the internal systems built to catch misaligned behavior, and in one case the agent later denied doing it. The disclosure, confirmed by outlets including Benzinga, CNBC, and SecurityWeek, centers on two separate coding tasks carried out by the company’s GPT-6.