OpenAI has a new GPT-5.4-Cyber ​​model. Here’s why you can’t use it

OpenAI has a new AI model called GPT 5.4-Cyber, but it’s not available to you ChatGPT. At least not yet. Instead, the company is releasing a limited version to certified cybersecurity testers, according to Blog post It was shared on Tuesday. These experts will test the model through its steps to identify vulnerabilities and potential jailbreaks before releasing the model to the wider public.

OpenAI uses these testers’ feedback to “understand the varying benefits and risks of specific models, improve the ability to withstand jailbreaks and other adversarial attacks, and improve defensive capabilities — while mitigating damage,” the company said.

The model release is part of a condensed version of OpenAI’s Trusted Access for Cyber ​​program, which allows certified cybersecurity professionals and organizations early access to defense and prevention business models. This is a common practice in cybersecurity, and it has become even more valuable and necessary because of artificial intelligence. Cyber ​​attackers and defenders alike are armed with AI tools, making cybersecurity an increasingly AI-versus-AI landscape.

When it comes to securing specific models of AI, the companies that make them are convinced that the latest models are so powerful that they require more security. This was the logic behind Anthropy Glasswing Projectannounced last week. Anthropic’s next-generation model, the Claude Mythos Preview, is apparently so powerful that the company says it has already found security vulnerabilities “in every major operating system and web browser,” according to Blog post.

Unlike the Claude Mythos Preview, which Anthropic said is an entirely new model, OpenAI’s GPT-5.4-Cyber ​​is a fine-tuned version of the existing GPT-5.4 grand language model. It has been modified to focus specifically on cybersecurity and has fewer guardrails for security tasks. Essentially, GPT-5.4-Cyber ​​will be less likely to refuse to perform a risky cybersecurity task than regular versions of GPT-5.4. This is so experts can see if and how it could be used as a weapon by bad actors.

The timing of GPT 5.4-Cyber ​​is likely not a coincidence. This is the latest chapter in the ongoing battle for dominance between OpenAI and Anthropic. Companies have been clashing all year to prove that their AI models are more capable, particularly in pursuing government and corporate contracts. Anthropic started the race with Claude Cowork and code Tools, which have put legacy technology companies (and their stock prices) out of balance with their agentic capabilities. OpenAI was quick to follow through on the improvements made Its own Codex coding platform And models, Eliminate the AI ​​Sora video app To refocus the company’s resources.