As I’ve previously written, Mythos Preview is a genuine inflection point for our industry. Anthropic has built something that can find vulnerabilities in code at a speed and scale no human team can match, and that will make the world’s software meaningfully more secure over time. We welcome that, and also recognize this is not a one-time event as capabilities like these become part of a broader wave of frontier model releases and the new baseline for an AI-first world.
At the same time, these same capabilities will increasingly be in the hands of attackers, who will use them to find and exploit vulnerabilities at the same AI speed. The pace of exploitation and the pace of fixing will both accelerate, which means that assuming vulnerabilities will always exist in production is not pessimism, it is just realism.
The question worth sitting with is not whether these advances change things (they do) but where the real enterprise risk lives in a world where AI capabilities like this are available to everyone.
Glasswing’s importance and how Netskope helps
Netskope is very proud to be part of Anthropic’s Project Glasswing, which has provided us access to Claude Mythos Preview. We will use this collaboration to identify and fix vulnerabilities in Netskope software, securing AI-rich environments. We will share our findings with the Glasswing coalition and with the security community at large, as we all benefit from what these frontier models are providing.
Netskope is the global inline inspection point for AI and all enterprise traffic for our customers, processing trillions of transactions for the world’s largest organizations. From this unique vantage point, a clear picture has come into focus. Among enterprises tracked by Netskope Threat Labs, the average enterprise organization saw the number of AI applications in use grow fivefold in the past year. That average organization tripled its AI user base; now manages 37 deployed AI agents; and sees 223 AI data policy violations per month.AI has become a net information-generation engine, and the risk is concentrated in what flows in and out of those applications every single day.
Defense in depth has never mattered more
Defending in this era, as in previous eras, is never about one layer. The architectures that will carry enterprises forward are the ones built for how work actually happens in 2026, with real-time governance and security that understands the language of the AI and cloud world. Every agent operating inside an enterprise must be zero-trusted, just like human users are in a true least-privilege modern architecture. Every transaction, regardless of whether it originates from a person, a device, or an AI agent, needs to be inspected, governed, and controlled in real time.
Ultimately, attackers want your data. That is a data, identity, and real-time transaction problem, in addition to a code problem. And the answer to that problem is open, converged platforms built for how work actually happens today, not tools designed for a perimeter that no longer exists.
We look forward to continued, fruitful collaboration with Anthropic. Earlier this month, we launched Netskope’s AgentSkope, our platform for deploying AI agents across security and networking workflows. As Michael Moore, Anthropic’s Head of Cybersecurity Products, said at the time: “Security teams need AI that can triage noise, connect signals across systems, and explain its reasoning to the analyst on the other end. Claude is built for that kind of work, and Netskope brings the platform, data, and SecOps expertise to apply it across security workflows. AgentSkope is a strong example of how the two can help teams keep pace with today’s threats.”
We also recently announced our integration with Anthropic’s Claude Compliance API, connecting Netskope’s data governance and compliance controls directly to Claude Enterprise usage across your organization.
For more:
Read my full perspective on Mythos
Read our Cloud and Threat Report 2026: the data that predicted this moment.
View the Netskope AI Index, the authoritative source on the enterprise use and adoption of AI worldwide.
Join Netskope’s AI in the Fast Lane roadshow, which is coming to a city near you: https://www.netskope.com/company/events.
