Anthropic Cyber Mission

Anthropic launched its Cyber Mission on October 8, and the most interesting line is an admission: finding vulnerabilities is easier than ever, but verifying, prioritizing and fixing them is still hard.

That comes from the lessons of Project Glasswing, where partners uncovered many vulnerabilities without a sufficient drop in cyber risk. More discovery did not translate into less exposure.

The response has two parts. The Critical Infrastructure Defense Program puts frontier models, on-site engineers and threat research in front of the teams that protect power grids, water systems, transportation networks and government systems. OSS Scanner gives open-source projects regular security scans from Anthropic's strongest models, for free.

My read: the AI labs are moving from selling capability to taking on a share of the responsibility for the risk that capability creates. The same models that can exploit software are being pointed at defending it, and the people doing the defending have had the least budget for decades.

For any executive running operational technology or depending on open-source code, the question changes. It is no longer whether AI can find your weaknesses. It is whether your team can fix them as fast as they are found.

Where is your organization's bottleneck today: finding vulnerabilities, or fixing them?