tech π€ OpenAI drops GPT-5.6-Cyber with fewer safety checks
OpenAI launched GPTβ5.6βCyber, a new model focused on vulnerability research and exploit development. It is being made available through Daybreak Red to firms for authorized security testing. Internal tests show it completes 95.0% of high-risk requests, much higher than GPTβ5.6 Sol's 1.5%. This model has already found high-severity flaws like CVE-2026-15903. Trusted partners like Cisco and Fortinet are now using it to close security gaps. π§