OpenAI has shared that it is working on the launch of a new cybersecurity model called GPT-5.5-Cyber, which will be available to “critical cybersecurity defenders” in the coming days, in order to protect businesses and infrastructure.
The company led by Sam Altman presented GPT-5.5 last Friday as its most intelligent model with improved capabilities that move towards “more intuitive computing”, allowing tasks to be carried out autonomously and understanding user intentions more accurately and quickly.
Just a week after its launch, the technology company has announced that it is “beginning the deployment” of a new version of this model, specialized in carrying out “cutting-edge cybersecurity” tasks.
This is the case of the new GPT-5.5-Cyber model, as Altman himself has shared in a publication on the social network
we’re starting rollout of GPT-5.5-Cyber, a frontier cybersecurity model, to critical cyber defenders in the next few days.
we will work with the entire ecosystem and the government to figure out trusted access for cyber; we want to rapidly help secure companies/infrastructure.
— Sam Altman (@sama) April 30, 2026
Although he has not shared more details about this new model and its specific characteristics for protection and cybersecurity, the manager has detailed that they will work with “the entire ecosystem and the government to determine reliable access for cybersecurity.” “We want to quickly help protect companies and infrastructure,” said Altman.
In this framework, it must be taken into account that its competitor Anthropic already presented its new Claude Mythos model in early April, capable of identifying high severity vulnerabilities in all major operating systems and web browsers. This model was unveiled under Project Glasswing, with the goal of “transforming cybersecurity.”
However, after testing its capabilities, it has been identified that beyond focusing on cybersecurity, it can also be a dangerous weapon in the wrong hands, as it can carry out attacks on its own against small businesses with weak defenses, as well as discover and exploit vulnerabilities autonomously.
Likewise, despite the fact that Claude Mythos was released in advance so that its capabilities are not used maliciously, a group of computer scientists also claimed to have gained unauthorized access to the model.