Endpoint Security Faces a New AI Reality
Mike Vizard speaks with Omer Singer, co-founder and CTO of Glow, during Techstrong TV’s Black Hat 2026 coverage. The conversation focuses on endpoint security for AI agents and why traditional endpoint detection and response may not be enough for the next wave of attacks. Singer says Glow is building a prevention-first endpoint platform that uses AI agents to control what runs on employee devices.

That shift is becoming urgent as organizations adopt AI tools across the workforce. Employees are using copilots, plugins, MCP servers, packages and other agentic tools to automate work. Security teams often lack full visibility into what is running on endpoints. That creates risk because these tools can access data, invoke actions and introduce new paths for attackers.
Prevention Moves Ahead of Detection
Singer argues that endpoint security for AI agents requires a new model. Detection and response worked when attackers moved more slowly. In an AI-driven threat environment, vulnerabilities can be discovered and exploited much faster. That makes prevention more important because security teams may not have enough time to detect, investigate and respond after the fact.

Glow’s platform is built around specialized AI agents that learn what good looks like inside an organization. They focus on the intent of the CISO and security team, then compare that intent with what is actually happening on endpoints. The goal is to reduce risk, remediate issues and keep devices aligned with approved policy without creating unnecessary friction for employees.
AI Tools Need Guardrails on the Endpoint
The discussion also looks at the balance between user flexibility and organizational control. Singer says security teams need smarter ways to enforce policy while still allowing employees to work. AI can help by supporting conversations with users, managing exceptions and applying more granular controls than older endpoint tools could support.

That matters because personal AI accounts, unapproved tools and shadow AI applications can spread quickly. Endpoint security for AI agents gives organizations a way to see those tools, understand their risk and guide employees toward safer alternatives. It also helps security teams support AI adoption instead of simply blocking it.
Open-Weight Models Add Urgency
Singer warns that open-weight models may soon approach the capabilities of frontier models being tested in labs. That could make advanced attack techniques more widely available. He says security leaders are already sensing urgency because software engineering and security work are changing quickly.

For Techstrong TV viewers, the takeaway is clear. Endpoint security has to evolve as AI agents become part of everyday enterprise work. Glow is positioning its platform as a way to help organizations prevent risk, manage AI adoption and prepare for machine-speed attacks.