OpenAI presented a new AI-based cyber tool to the US government and allies - Axios
Kyiv • UNN
OpenAI has developed a specialized model for finding vulnerabilities in government systems. Only verified entities will have access to the powerful version.
OpenAI has begun closed briefings for US federal agencies, individual state governments, and representatives of the Five Eyes alliance countries regarding its new cybersecurity AI product, GPT-5.4-Cyber. This is reported by Axios, according to UNN.
Details
According to the publication, the company recently held a closed event in Washington for approximately fifty government officials responsible for cybersecurity. During the meeting, OpenAI demonstrated the capabilities of the new model.
This refers to a specialized tool focused on tasks in the field of cyber defense. Access to the system is controlled through the Trusted Access program, which involves user verification.
As Axios notes, the company is launching the product in two formats. One version will have additional safeguards for broader use, while a more powerful version will only be available to verified defense and cybersecurity entities.
According to officials, the new tool can help identify vulnerabilities in outdated government information systems.
At the same time, the publication adds that OpenAI's competitor, Anthropic, is also testing a similar model, but with limited access due to security risks.
Britain believes 100 countries have cyberattack technologies22.04.26, 12:15 • 1908 views