AI Pulse · Frontier Labs Watch

Anthropic is widening the defensive use of Mythos 5, its AI capability for security work. The company says it is embedding the model into partner security tools, adding vulnerability scanning to Claude Security for enterprise customers, and launching a $35 million Defender Advantage Fund. That fund will provide credits to organisations working to patch vulnerabilities in open-source software.

For a business owner, the practical point is that AI security is moving beyond writing alerts or summarising incident reports. Vendors are starting to place AI directly into the workflow of finding weaknesses, prioritising them, and helping teams respond. That can matter most for companies with lean IT or security teams, where known vulnerabilities can sit unpatched simply because nobody has the time to sort through the backlog.

The important question is not whether an AI tool can scan for issues. Many tools already do parts of that. It is whether your team has a clear process for deciding what gets fixed first, who owns the work, and how you verify the patch did not create another problem. This is the kind of thing we build into an AI command centre, connecting AI activity to the operating decisions and controls around it.