Glow, a Palo Alto-based cybersecurity startup founded in 2025, emerged from stealth on July 22, 2026 with a $180 million all-equity Series A at a $1.2 billion valuation. The round drew a syndicate of nine firms: Sequoia Capital, Cyberstarts, Greenoaks, Redpoint Ventures, Index Ventures, Swish Ventures, Lux Capital, Operator Collective, and Holly Ventures TechCrunch.

The company is building an endpoint security platform built around specialized AI agents that map enterprise environments, assess risk in real time, and enforce security policies. Glow uses models from Anthropic and Google's Gemini, accessed through Amazon Bedrock rather than self-hosted inference. The choice to build on managed foundation-model APIs rather than train proprietary models from scratch is notable: it lets a startup concentrate on the agentic orchestration and policy-enforcement layer while relying on frontier-lab partners for the underlying reasoning capability.

Glow's founding team brings deep lineage from both consumer-platform engineering and enterprise security. CEO Roi Tiger was a vice president of engineering at Meta. Co-founders include Omer Singer, formerly head of cybersecurity strategy at Snowflake; Ophir Arie, formerly VP of R&D at Claroty; and Arnon Joseph, another former Meta engineering leader. COO Emily Heath served as CISO at United Airlines and DocuSign, sat on the board of Wiz, and was a partner at Cyberstarts before joining Glow. That bench spans the builder-operator and buyer-practitioner perspectives in roughly equal measure.

The company's press release, datelined Tel Aviv, was titled "Glow Emerges From Stealth With $180 Million to Reinvent Endpoint Security in the AI Era" Glow. TechCrunch obtained its story through a direct interview with Tiger, giving the outlet first-hand access to the announcement.

Glow says it already has paying customers across healthcare, retail, and financial services, with typical deployments spanning tens of thousands of employee devices in global organizations. The company declined to disclose customer names or headcounts.

The competitive field is crowded. Glow enters an endpoint security market where CrowdStrike, Microsoft, SentinelOne, and Palo Alto Networks are entrenched, each with their own investments in AI-assisted detection and response. What differentiates Glow's approach, at least architecturally, is the framing of AI agents as autonomous operators within the security stack rather than as embedded ML features bolted onto an EDR pipeline. The agents are designed to continuously discover and map the environment, evaluate risk posture in real time, and enforce policy decisions, all without a human in the loop on every action.

Whether that autonomous-enforcement model holds up under the operational and regulatory constraints that large enterprises impose remains an open question. Endpoint security buyers in regulated industries, where Glow already claims traction, tend to demand granular audit trails and human-validated responses for at least high-severity actions. An agentic system that can enforce policy autonomously is only as trustworthy as its hallucination rate, its policy-interpretation fidelity, and the latency of its decision loop relative to an active threat. None of those characteristics are visible from outside the company today.

The $1.2 billion post-money valuation at the Series A stage also warrants scrutiny. Enterprise-security unicorns at emergence are not unprecedented in this cycle, but the valuation implies revenue multiples and growth trajectories that Glow has not yet substantiated publicly. The company's willingness to forgo structured-equity instruments, taking all-equity capital, signals confidence from both founders and investors that the cap table can absorb the dilution without downside protection. It also concentrates risk: if the agentic-security thesis does not translate into durable enterprise contracts, there is no ratchet cushion.

The broader context here is that endpoint security is being re-architected around AI at multiple layers simultaneously. CrowdStrike and SentinelOne have integrated generative models into investigation workflows. Microsoft's Defender line leverages its own model infrastructure. Palo Alto Networks has folded AI capabilities across its Cortex platform. Glow's bet is that incumbents are constrained by legacy architectures optimized for signature-based and behavioral detection, and that a clean-slate, agent-first design can outperform a retrofitted stack. That is a plausible thesis. It is also the same thesis that every security startup of the past decade has articulated in one form or another.

What may work in Glow's favor is timing. Foundation-model capabilities have crossed a threshold where autonomous multi-step reasoning over complex environments is no longer theoretical. The Bedrock dependency means Glow inherits the latency, cost, and data-handling characteristics of its model providers, but it also means the platform improves as those models improve, without retraining. For a startup that cannot afford to train frontier models, that is a reasonable architectural bet.

The founding team's pedigree, the capital base, and the agentic architectural approach give Glow a credible entry point. The execution questions, the ones that will determine whether a $1.2 billion valuation is earned or merely raised, are whether autonomous policy enforcement can meet the bar in regulated enterprises, whether the platform's risk-assessment accuracy holds at scale, and whether customers will pay a premium for an agent-native approach over the AI-enhanced incumbents they already run. Those answers will come from deployment data Glow has not yet shared.