OpenAI roll-out of GPS-5.6-Cyber model to enhance advanced cybersecurity mission capabilities

On 11 August, OpenAI announced the launch of GPT-5.6-Cyber and the expansion of the Cyber Security Plan Daybreak to provide an AI model that is better suited to the advanced security of the approved cybersecurity defence personnel. OpenAI expressed the hope that the forward AI capability would be handed over to credible defence personnel before the attackers deployed an autonomous cyber attack on a large scale. Daybreak is divided into two levels: Daybreak Blue provides GPS-5.6 Sol access, with no system-level network security protection restrictions; Daybreak Red provides GPS-5.6-Cyber access, which is used to exploit loopholes and advanced gap studies. Tests showed that GPT-5.6-Cyber responded to 95 per cent of requests for advanced cybersecurity missions, including gap-use chain development, identification bypasses and enhanced privileges. OpenAI stated that high-risk capabilities would be open only to authorized users with additional controls and controls。