IBM plans to integrate Anthropic’s Claude Mythos artificial intelligence model into its Concert Protect security software, allowing customers to identify previously unknown vulnerabilities in their code and assess whether they can be exploited.
The company said on Monday that it was targeting early 2027 for the rollout, although access to Mythos itself remains restricted to vetted organisations.
The integration will enable Concert Protect to scan code selected by customers for security flaws that conventional scanning tools may overlook.
The software will then assess whether the identified vulnerabilities could be exploited by attackers within customers’ systems and provide recommended fixes for developers to review and approve.
IBM and Red Hat tools will also generate proposed fixes for identified vulnerabilities, which developers must review and approve before implementation. However, IBM cautioned that its plans could change.
The company is also exploring partnerships beyond Anthropic. In June, it joined OpenAI’s cybersecurity programme to develop a similar service.
“The bottleneck is no longer finding vulnerabilities. It is determining which can actually be exploited and fixing them before attackers do,” IBM wrote.
The challenge is reflected in findings from Project Glasswing, Anthropic’s restricted programme for Claude Mythos.
The initiative identified more than 10,000 potentially serious security flaws in its first month, but only 97 had been patched at the time.
IBM said its previous work with Mythos through Project Glasswing was limited to internal research.
