Glow: The AI Endpoint Security Startup Sequoia Backed — How It Works

Overview

Today, a startup called Glow stepped out of stealth with one-hundred-eighty million dollars led by Sequoia, at a one-point-two-billion-dollar valuation — to fix a problem that barely existed two years ago: the A-I now running on your employees' laptops. Three things tonight — the technology behind Glow, how it actually

In this show

• Today, a startup called Glow stepped out of stealth with one-hundred-eighty million dollars led by Sequoia, at a one-point-two-billion-dollar valuation — to fix a problem that barely existed two years ago: the A-I now running on your employees' laptops. Three things tonight — the technology behind Glow, how it actually works, and why it matters. With me, an Endpoint Security Architect and an A-I Threat Researcher.
• Start with the shift. For a decade, security chased the software as everything moved to the cloud and to SaaS. Now something new has happened: A-I has landed on the endpoint itself. Copilots, autonomous agents, and A-I developer tools run directly on the laptop. In a single year, employee use of A-I tools on company devices jumped from fifteen percent to forty-five percent — far faster than any security team could watch. The device became the new frontier.
• And that frontier is dangerous. Attackers now wield generative A-I at machine speed — automating phishing, writing malware, and hunting for software flaws faster than any human could. After Anthropic's Mythos model showed how capably A-I can find and exploit vulnerabilities, that fear sharpened. The endpoint — the very place these new A-I tools live — has become the softest, richest target in the enterprise.
• Here is why the old playbook struggles. Traditional endpoint tools — the E-D-R systems from CrowdStrike and SentinelOne — are built to DETECT a threat and RESPOND after it appears. They are reactive by design. Against attacks moving at A-I speed, and a brand-new attack surface made of A-I agents, waiting for the alarm to sound is already waiting too long. You have to stop the risky thing before it ever runs.
• That is Glow's core idea: prevention-first. Instead of catching threats after they land, stop risky software, rogue A-I agents, and dangerous developer tools from ENTERING the device in the first place. As Glow's chief executive framed it — prevention was always the right answer in security; it just never worked at enterprise scale without blocking the business. Their bet is that A-I finally makes prevention practical.
• So how does it actually work? Glow deploys specialized A-I agents onto each device. They continuously MAP everything running — every application, every A-I agent, every developer tool. They ASSESS the risk of each one in real time. And through a context-and-reasoning engine, they AUTONOMOUSLY enforce policy — deciding, live, what is allowed to stay and what gets removed, without a human weighing in on every single call.
• Concretely, that means blocking a malicious N-P-M code package before a developer ever installs it, catching an A-I agent trying to pull down dangerous software, and flagging any laptop where the security tools have gone missing or degraded. Under the hood, Glow leans on frontier models from Anthropic and Google Gemini, run through Amazon Bedrock, wrapped in its own software that gives those models real enterprise context.
• The promise is prevention WITHOUT blocking the business — security that keeps pace with A-I adoption instead of fighting it. And the team is heavyweight: founders drawn from Meta, Snowflake, and Claroty, and a chief operating officer who was a chief information security officer at United Airlines. Backed by Sequoia and Cyberstarts, Glow already protects tens of thousands of devices across healthcare, retail, and finance.
• Three takeaways. First — A-I has moved onto the endpoint, an attack surface that grew from fifteen to forty-five percent of devices in just one year. Second — Glow is prevention-first: A-I agents that map every device, score risk in real time, and enforce policy autonomously, blocking danger before it runs instead of detecting it after. Third — it is a one-point-two-billion-dollar bet, led by Sequoia, that security must now move at A-I speed. Thank you both.