OpenAI is getting ready to launch a brand new frontier cybersecurity mannequin, GPT-5.5-Cyber. CEO Sam Altman mentioned the mannequin is not going to be obtainable to most people, however might be first rolled out to a choose group of trusted “cyber defenders” to ensure that establishments to shore up their cyberdefenses.
The restricted rollout will happen “within the subsequent few days,” Altman said on X. “We’ll work with the whole ecosystem and the federal government to determine trusted entry for Cyber.”
It’s not clear who will get entry to the mannequin first, although previous “trusted entry” schemes concerned vetted professionals and establishments. Particulars of the mannequin and its capabilities are additionally unclear; OpenAI has not launched any technical particulars or specs. The identify signifies it’s a specialised model of the recently released GPT-5.5which it known as its “smartest and most intuitive to make use of mannequin but.”
The staggered rollout is a part of a rising development within the AI business of corporations branding their prime fashions too harmful for public launch as a result of their potential for misuse. OpenAI has staggered the discharge of earlier cybersecurity-focused fashions, along with its new purpose-built life sciences model GPT-Rosalindwhich is meant to help biology analysis and drug discovery. This month, Anthropic adopted an identical playbook with Claude Mythos, although with a lot larger fanfare, and it bungled the model’s secure release in embarrassing methods.
The White Home has taken a eager curiosity in Mythos’ rollout, regardless of lingering tensions with Anthropic after its combat with the Pentagon. It has lately opposed plans to broaden entry to Mythos additional, according to The Wall Road Journal. The report’s unnamed White Home officers cite each cybersecurity issues related to extra individuals accessing Mythos and worries that elevated demand would hamper the federal government’s capacity to make the most of the system as causes for the pushback.
