IBM announced on Monday that it will incorporate Anthropic's Claude Mythos into Concert Protect, its security software suite. The integration targets early 2027 and would enable organisations to identify previously unknown vulnerabilities in their own codebases, assess exploitability, and generate fixes for developer review and approval. Mythos itself continues to operate under strict access controls, available only to vetted partners.

The model would examine code selected by customers to uncover zero-day vulnerabilities and flaws that conventional security scanners typically overlook. Once identified, Concert Protect would evaluate whether each vulnerability could actually be leveraged against the customer's infrastructure. Proposed patches generated by IBM and Red Hat tools would then require developer sign-off before deployment. IBM noted that these plans remain subject to change and are not contingent on a single AI provider—the company joined OpenAI's cybersecurity programme in June to pursue comparable capabilities.

The bottleneck is no longer finding vulnerabilities. It is determining which can actually be exploited and fixing them before attackers do.

IBM

The challenge IBM identifies reflects a genuine operational gap. During the first month of Project Glasswing, Anthropic's restricted Mythos programme, the model flagged over 10,000 potential serious vulnerabilities, yet only 97 had received patches by that point.

This announcement represents IBM's initial commercial move toward Mythos deployment. Prior work through Glasswing was confined to internal research. The approach aligns with Anthropic's broader strategy of embedding Mythos within partner platforms, where end users receive vulnerability findings and remediation suggestions rather than direct access to the model itself.

Access to Mythos remains highly restricted. Anthropic temporarily disabled the model in June to adhere to US export control regulations, and the White House reportedly now has authority over approvals for its partner programme. Anthropic released Mythos 5.2 on Monday, maintaining the same access limitations for vetted organisations only.

Source: The Next Web