Following the release of Anthropic Mythos, OpenAI has announced GPT-5.4-Cyber, a new specialized model designed specifically for cyber defense operations.
Key Features and Training
The GPT-5.4-Cyber model is trained to minimize false refusals during legitimate security tasks. It excels in high-level security functions, including:
- Binary analysis
- Reverse engineering
- Malware identification
- Vulnerability detection (even without source code access)
Restricted Access
To ensure responsible use, OpenAI is implementing a verification process. Initially, access to the technology will only be granted to verified researchers and organizations that complete a rigorous identity check.
