OpenAI Expands Cyber Push With GPT-5.4-Cyber Launch
GPT-5.4-Cyber is aimed at vetted defenders working on malware analysis, vulnerability research, and reverse engineering under OpenAI’s Trusted Access for Cyber framework.
Topics
OpenAI has launched GPT-5.4-Cyber, a cybersecurity-focused version of GPT-5.4 designed for vetted defenders to analyze malware, find vulnerabilities, and strengthen software security through its Trusted Access for Cyber (TAC) program.
The release comes a week after Anthropic introduced Claude Mythos Preview under Project Glasswing as competition to build AI tools for cyber defense intensifies.
Anthropic said Mythos would be used to help secure critical software and systems.
OpenAI said GPT-5.4-Cyber is a refined version of GPT-5.4 with lower refusal thresholds for legitimate cybersecurity tasks. That allows trusted users to work on more advanced security problems, including large codebase analysis, vulnerability research, malware analysis, and reverse engineering of binaries when source code is unavailable.
One of the model’s key capabilities is binary reverse engineering, which lets analysts inspect compiled software to understand behavior, detect threats, and assess security weaknesses without access to source code.
Reuters reported that OpenAI is initially limiting access to vetted security professionals, organizations, and researchers because of the model’s more flexible functionality.
OpenAI, which introduced the TAC program in February, on Tuesday, 14 April, said it was expanding the framework to thousands of verified individual defenders and hundreds of teams.
New access tiers are intended to match model capability with verification level and defensive use case.
