OpenAI has introduced a new version of ChatGPT, named GPT-5.4-Cyber, designed specifically for cybersecurity applications. This model enhances the ability of professionals to identify and counter potential cyber threats. The company stated that this version has been optimized for greater hacking capabilities and features fewer restrictions on its ability to explore vulnerabilities.
The launch follows the release of Anthropic's Claude Mythos, which raised concerns about the security implications of advanced AI technologies. OpenAI plans to restrict access to GPT-5.4-Cyber to vetted organizations, focusing on those tasked with protecting critical infrastructure and public services. The company aims to ensure that these tools are available to those with legitimate needs while implementing measures to prevent misuse.
Automated verification systems are being developed to assess access requests, enhancing the process of granting permissions based on trust and accountability rather than subjective decisions. OpenAI emphasizes the need to empower legitimate defenders to secure their systems against emerging threats, facilitating a broader rollout as cybersecurity professionals prepare for the advent of more powerful AI models.