OpenAI has introduced the GPT-5.4-Cyber model designed specifically for cybersecurity professionals. This version features reduced refusal rates, allowing security researchers to perform bug hunting and defensive programming without being blocked by safety filters.

GPT-5.4-Cyber offers advanced binary reverse engineering capabilities, enabling analysis of compiled software for malware and vulnerabilities even without access to source code. Due to its dual-use nature, access is strictly controlled via the Trusted Access for Cyber program, requiring identity verification and corporate manager approval.

Source: OpenAI