CrowdStrike Launches Project QuiltWorks and Extends Falcon Platform Into Google's Agent Cloud Ecosystem

CrowdStrike announced two significant moves into agentic AI security this week: the launch of Project QuiltWorks, a cross-industry vulnerability discovery initiative, and its designation as a launch partner for Google’s Agent Cloud ecosystem with Falcon platform integration for runtime security of autonomous agents.

Project QuiltWorks combines frontier AI models with CrowdStrike’s Falcon Spotlight vulnerability discovery to accelerate finding and fixing software flaws before attackers reach them. Initial participants include OpenAI, Accenture, EY, IBM Cybersecurity Services, and Kroll, according to CRN.

The Scale of the Problem

CrowdStrike Chief Business Officer Daniel Bernard told CRN that one QuiltWorks participant has already found 45 million vulnerabilities using the combined capabilities. “We’re going to do more patching in the next six months to 12 months than has ever been done on planet Earth before,” Bernard said. “Before it was Patch Tuesday, once a month. Now it’s, ‘Patch every day, all the time.’ That’s what this new world looks like.”

The initiative was triggered by Anthropic’s Project Glasswing disclosure earlier this month, which demonstrated that AI models like Claude Mythos could exponentially increase vulnerability discovery. Bernard said demand surged in the two weeks following that announcement: “All of a sudden, there was a deluge … of demand and interest around, ‘What do all these models mean? And are we ready for these models?’”

Harness Over Model

Bernard emphasized that Project QuiltWorks is intentionally model-agnostic. “The harness actually matters now more than the model,” he told CRN. CrowdStrike is one of only two pure-play cybersecurity vendors participating in both Anthropic’s Project Glasswing and OpenAI’s Trusted Access for Cyber program, though the Mythos model itself cannot be used within QuiltWorks due to Glasswing restrictions.

Google Agent Cloud Integration

Separately, CrowdStrike was named a Google Agent Cloud ecosystem launch partner, extending the Falcon platform to provide guardrails, visibility, and runtime control as agentic AI systems move from experimentation to production, according to PNI News. Google also named CrowdStrike its 2026 Security Partner of the Year for Infrastructure Protection for the second consecutive year.

The Cloud Detection and Response extension eliminates log batch processing delays that typically take 15 minutes or more to surface a detection. CrowdStrike’s real-time detection engine analyzes cloud activity as it happens using event streaming, according to CXO Digital Pulse. The expansion also extends Falcon to regional Google Cloud infrastructure for data sovereignty compliance.

“As organizations operationalize agentic AI, they need real-time protection, complete visibility, and the ability to stop breaches across cloud, identity, and AI attack surfaces,” Bernard said, according to PNI News.

The Runtime Security Bet

CrowdStrike’s dual announcements represent a bet that the security layer for autonomous agents will be built at the endpoint and runtime level, not through posture scanning or model-level restrictions alone. The combination of QuiltWorks (finding vulnerabilities before agents do) and Agent Cloud integration (monitoring agents as they execute) positions CrowdStrike on both sides of the agent security equation. With Anthropic’s Mythos breach this week showing that restricted access models leak anyway, runtime monitoring may prove more durable than access control as a defense strategy.