- OpenAI introduces GPT‑5.5‑Cyber following uproar round Anthropic Mythos
- It is a modest improve targeted on permissive cybersecurity duties like vuln triage and malware evaluation
- Entry is proscribed to vetted groups within the Trusted Entry for Cyber (TAC) program, in contrast to Anthropic’s extra restricted Mythos Preview
OpenAI has launched GPT-5.5-Cyber, an upgraded cybersecurity mannequin trying to take among the shine off Anthropic’s Mythos Preview launch.
Coming lower than a month after the launch of GPT-5.4-Cyber, this isn’t a serious improve by any means, and customers shouldn’t count on many adjustments, OpenAI defined.
as an alternative, customers ought to count on a mannequin educated to be extra permissive in the case of cybersecurity duties, making it simpler to make use of for issues like vulnerability identification, triage, patch validation, and malware evaluation.
Newest Movies From
You might like
Competing with Mythos
“GPT‑5.5‑Cyber lets a smaller set of companions research superior workflows the place specialised entry conduct could matter,” OpenAI stated in a weblog submit.
“The cyber protection ecosystem is broad, and GPT‑5.5 and GPT‑5.5‑Cyber play completely different roles in assembly the wants of organizations and researchers throughout it, relying on the duty, the setting, and the safeguards round how the mannequin is used. For many groups, GPT‑5.5 with TAC is our strongest broadly helpful mannequin for legit defensive work, with sturdy safeguards in opposition to misuse.”
As with the earlier model, this version will solely be given solely to vetted cybersecurity groups. Nonetheless, in contrast to its key competitor – Mythos, which was given to solely a handful of firms, OpenAI’s mannequin will likely be supplied to a broader set of customers, members of the Trusted Entry for Cyber (TAC) program.
Again when it launched GPT-5.4-Cyber, OpenAI stated it was scaling TAC to “hundreds of verified particular person defenders and a whole bunch of groups liable for defending essential software program.”
Anthropic first disclosed Venture Glasswing in early April 2026, saying that the AI mannequin Mythos Preview was too highly effective to be given freely. Apparently, it was capable of floor decades-old vulnerabilities in among the most widely-used working techniques in existence, and chain them collectively to create working exploits.
The most effective antivirus for all budgets
Our prime picks, primarily based on real-world testing and comparisons
Observe TechRadar on Google Information and add us as a most popular supply to get our skilled information, critiques, and opinion in your feeds.
