OpenAI began briefing United States federal agencies and Five Eyes intelligence allies on a new artificial intelligence model designed specifically for cyber defense.
The intelligence-sharing network comprises the U.S., Britain, Canada, Australia, and New Zealand. Briefings held in Washington focused on GPT-5.4-Cyber, a variant of the flagship model fine-tuned for defensive cybersecurity operations.
The model features binary reverse engineering, a process allowing security teams to inspect compiled software for malware risks and flaws without the original source code. OpenAI distributes access through the Trusted Access for Cyber program.
This initiative provides verified security professionals and government agencies controlled use of the model. The company restricts broad release due to the dual-use nature of the technology. Frontier models possess the capability to find vulnerabilities and generate exploits, software tools taking advantage of system weaknesses.
Security experts note that highly capable AI models are increasingly treated as controlled infrastructure rather than open commercial products.
Source:
Cover Photo by Pixabay