April 14 (Reuters) – OpenAI on Tuesday unveiled GPT-5.4-Cyber, a variant of its newest flagship mannequin fine-tuned particularly for defensive cybersecurity work, following rival Anthropic’s announcement of frontier AI mannequin Mythos.
Mythos, introduced on April 7, is being deployed as a part of Anthropic’s “Challenge Glasswing”, a managed initiative underneath which choose organizations are permitted to make use of the unreleased Claude Mythos Preview mannequin for defensive cybersecurity functions. It has discovered “1000’s” of main vulnerabilities in working programs, internet browsers and different software program.
OpenAI, creator of standard chatbot ChatGPT, stated that GPT-5.4-Cyber will initially be rolled out on a restricted foundation to vetted safety distributors, organizations and researchers due to its extra permissive design.
The corporate can also be increasing its Trusted Entry for Cyber program to 1000’s of verified particular person defenders and tons of of groups defending vital software program, it stated in a submit on its web site.
OpenAI is including new tiers to its TAC program, which was launched in February, with larger ranges of verification unlocking extra highly effective capabilities.
Customers accredited for the very best tier will achieve entry to GPT-5.4-Cyber, which has fewer restrictions on delicate cybersecurity duties reminiscent of vulnerability analysis and evaluation.
