OpenAI launched GPT-5.4-Cyber, a fine-tuned GPT-5.4 variant that relaxes guardrails and adds capabilities like binary reverse engineering to help defenders analyze vulnerabilities and malicious behavior. The model will be distributed iteratively through an expanded Trusted Access for Cyber program for thousands of verified defenders and vetted security teams, contrasting with Anthropic’s restricted Claude Mythos rollout. #GPT54Cyber #ClaudeMythos
Keypoints
- OpenAI is scaling Trusted Access for Cyber to thousands of verified defenders and hundreds of security teams.
- GPT-5.4-Cyber relaxes normal guardrails to support legitimate cybersecurity work and introduces binary reverse engineering.
- Access is being offered iteratively to vetted vendors, organizations, and researchers, with individuals applying via chatgpt.com/cyber.
- OpenAI frames the effort around democratized access, iterative deployment, and ecosystem resilience, and highlights tools like Codex Security.
- The approach contrasts with Anthropic’s restricted Project Glasswing distribution of Claude Mythos amid dual-use concerns.