OpenAI introduces a new cyber model amid fears of AI cyberattacks

OpenAI introduced GPT-5.6-Cyber, a more permissive version of its model for vetted defenders. The move follows delays of its Astra model due to critical hacking abilities during safety testing. OpenAI expanded its Daybreak program with two tiers: Daybreak Blue (GPT-5.6 Sol without cyber guardrails) and Daybreak Red (GPT-5.6-Cyber). Accenture, IBM, CrowdStrike, Cisco and Palo Alto Networks can integrate the models into security products.