OpenAI Counters Anthropic’s Mythos With GPT-5.4-Cyber Launch
Similar to Mythos, GPT-5.4-Cyber will initially be rolled out on a limited basis to vetted security vendors, organizations, and researchers.
News
- AI Dispatch | 10–16 April 2026
- UAE Firms Shift to AI Messaging as WhatsApp Overtakes SMS
- OpenAI Counters Anthropic’s Mythos With GPT-5.4-Cyber Launch
- Saudi Arabia Tops Global Digital Readiness Index, Overtakes Finland and Germany
- Egypt to Roll Out Child-Friendly SIM Cards by Mid-2026
- A 50-Point Gap: How Experts and the Public See AI Differently
[Image source: Pankaj Kirdatt/MITSMR Middle East]
OpenAI has announced GPT-5.4-Cyber, a variant of its latest flagship model fine-tuned specifically for defensive cybersecurity, following competitor Anthropic’s announcement of Mythos last week.
Mythos, announced on April 7, is being deployed as part of Anthropic’s Project Glasswing, a controlled initiative under which select organizations are permitted to use the unreleased Claude Mythos Preview model for defensive cybersecurity purposes.
Applying a similar logic, OpenAI said that GPT-5.4-Cyber will initially be rolled out on a limited basis to vetted security vendors, organizations, and researchers. These testers will put the “cyber-permissive” model through its paces to identify gaps and potential jailbreaks before it is released to wider users.
“We want to empower defenders by giving broad access to frontier capabilities, including models which have been tailor-made for cybersecurity,” wrote OpenAI. “This is a version of GPT‑5.4, which lowers the refusal boundary for legitimate cybersecurity work and enables new capabilities for advanced defensive workflows.”
Early access to technological developments for defence and security is a common practice. With adversaries armed with AI tools today, the rate, scale, and intensity of cyber incidents have accelerated. Hence, equally powerful defence systems need to be put in place.
The OpenAI release is part of a ramped-up version of the company’s Trusted Access for Cyber (TAC) program, which it launched in February to build a network of users who have completed OpenAI’s automated identity verification process, including a government ID check.
However, not every TAC-approved user will immediately get access to the model. Those at the topmost tiers of the program will have access to GPT-5.4-Cyber at present. They will require further authentication to be verified as “legitimate cyber defenders.”
Unlike Mythos Preview, which, according to Anthropic, is an entirely new model, GPT-5.4-Cyber is a version of its existing GPT-5.4 model. It’s been fine-tuned to be cybersecurity-focused. The model allows queries related to security tasks, so it won’t refuse to perform a risky task, unlike versions available to the wider public.
The launch timing is just the latest chapter in the ongoing battle for AI dominance between the two companies. Both OpenAI and Anthropic have been in a tiff trying to prove their capabilities through model performance, funding, and expansion through government and enterprise deals.
