IBM plans to add Claude Mythos code scanning to Concert Protect
Developers would review proposed fixes before implementation, while access to Anthropic’s model would remain restricted to vetted organizations.
IBM is aiming for early 2027 to introduce code-scanning capabilities powered by Anthropic’s Claude Mythos in its Concert Protect security software, the company said. Customers would be able to look for unknown vulnerabilities in their own code, though IBM warned the plans could change.
The proposed workflow would use Mythos to examine customer-selected code, including for weaknesses conventional scanners fail to detect. Concert Protect would evaluate whether those vulnerabilities could actually be exploited in the customer’s systems.
Tools supplied by IBM and Red Hat would then generate suggested repairs. Developers would have to approve those changes after reviewing them, rather than having fixes applied without their authorization.
According to IBM, its previous work with Mythos under Anthropic’s restricted Project Glasswing program was for internal research. The Concert Protect plan represents an initial step toward selling customers access to its capabilities. Anthropic intends to deliver findings and fixes through partner products without giving users the model itself.
Glasswing flagged more than 10,000 potentially serious vulnerabilities during its first month; only 97 had been patched at that stage. Mythos remains restricted to vetted organizations. IBM also entered OpenAI’s cyber program in June for a comparable service.
