CrowdStrike used its three-day Fal.Con 2026 conference in Las Vegas to lay out a broad agentic security stack, spanning a new frontier AI research lab, runtime protection for AI agents on the endpoint, and identity infrastructure built specifically for autonomous systems.

Taken together, the announcements reflect a shift in enterprise security priorities as AI agents move from pilot projects into production, executing tasks with the access privileges of the humans they serve.

Day one: automating AI risk assessment at machine speed

CrowdStrike opened the conference by launching Falcon IQ, which operationalises Project QuiltWorks, the company’s AI-driven vulnerability discovery and remediation coalition. Powered by NVIDIA Nemotron open models and Charlotte AI AgentWorks, Falcon IQ uses more than 50 AI agents to automate assessment, prioritisation and remediation workflows.

“Falcon IQ changes the economics of securing frontier AI risk. Agents do the heavy lifting, partners scale delivery, and organizations gain protection in a fraction of the time and cost,” said Daniel Bernard, chief business officer at CrowdStrike.

Project QuiltWorks was also expanded to ingest real-time data from a dozen technology partners, including Netskope, Rubrik and Zscaler, into Falcon Next-Gen SIEM, while the Falcon platform became available on Google Cloud infrastructure.

Day two: a research lab built for AI safety

CrowdStrike’s most significant announcement came on day two, with the launch of its Cyber Superintelligence Lab, which it describes as the first frontier AI research organisation built for cyberdefence and AI safety. The Lab unites the company’s AI researchers, offensive operators and incident responders, and is led by Dr Bartley Richardson, chief AI and autonomous systems officer.

“Security is how AI scales. The synthesis of our data moat and cybersecurity’s very best AI talent advantages defenders like never before,” said George Kurtz, chief executive officer and founder of CrowdStrike.

The Lab’s first output is SafeMind, a family of security models developed with NVIDIA that pairs an offensive model, Red Tempest, against a defensive counterpart, Blue Solano, in a continuous loop designed to find and close vulnerabilities. CrowdStrike also introduced Falcon Guardian, a new AI Detection and Response solution that extends runtime enforcement to the endpoint, where AI agents execute, and across cloud, SaaS and browser environments.

Day three: identity and platform partnerships for the agentic era

On the final day, CrowdStrike introduced its Agentic Identity Provider, designed to issue cryptographically verifiable identities to AI agents, broker short-lived and tightly scoped access, and attribute agent actions back to the human or system they act on behalf of. The company also launched Real-Time Supply Chain Attack Protection, which blocks malicious open-source npm and PyPI packages at the endpoint before their code can execute, and unveiled the next evolution of its agentic security operations centre, enabling coordinated multi-agent investigations across endpoint, identity, SaaS, cloud and network simultaneously.

CrowdStrike closed the conference with a series of ecosystem announcements, including bringing the Falcon platform to the Anthropic Claude Marketplace and expanding its partnership with OpenAI to extend Falcon Guardian’s runtime security to supported Codex agents.

• Cyber Superintelligence Lab: first frontier AI research organisation built for cyberdefence
• SafeMind: paired offensive and defensive AI models built with NVIDIA
• Falcon Guardian: runtime AI agent protection from endpoint to cloud, SaaS and browser
• Agentic Identity Provider: verifiable, short-lived identities for AI agents
• Falcon platform now available on Google Cloud and through the Anthropic Claude Marketplace

The announcements come as enterprises across Asia Pacific accelerate agentic AI adoption even as regional governments, including Singapore, move to formalise AI governance frameworks, putting fresh pressure on security vendors to demonstrate that autonomous agents can be run safely at scale.