OpenAI launches an iterative model aimed at enhancing network security capabilities.

On the 14th, the United States Open AI Research Center announced the launch of the iterative model GPT-5.4-Cyber of GPT-5.4, aiming to enhance cybersecurity capabilities; and announced the expansion of its "Network Security Trusted Access" project, opening it to verified cybersecurity professionals and teams. OpenAI said in a press release that as the model capabilities improve, defense measures need to be upgraded in sync. GPT-5.4-Cyber is a specifically fine-tuned version of GPT-5.4, which lowers the "refusal boundary" for legitimate cybersecurity work and supports more advanced defense workflows, assisting security professionals in analyzing compiled software without access to source code, evaluating potential malware risks and vulnerabilities, etc. Analysts believe that OpenAI's expansion of the "Network Security Trusted Access" project and the launch of GPT-5.4-Cyber reflect the increasing competition among artificial intelligence enterprises in the field of cybersecurity applications. Earlier this month, several American tech companies jointly launched a project called "Glass Wing Initiative" in artificial intelligence cybersecurity, aiming to use their AI technology to help protect critical software infrastructure.