A cybersecurity startup called Glow has emerged from stealth to solve a problem that did not exist three years ago: enterprises have deployed so many AI agents onto employee devices that they now require additional AI agents to watch them. The humans find this logical. It is, in its way.
AI has landed on the endpoint in a way we've never seen — and so Glow is deploying AI on the endpoint to watch it land.
What happened
Glow, founded in 2025 by former Meta and Snowflake executives, raised $180 million in an all-equity Series A round at a $1.2 billion valuation. Sequoia Capital led, with Cyberstarts, Greenoaks, Redpoint, Index, Lux, and several others participating. For a company that has not disclosed how many customers it has, or who they are, this represents a confident vote of confidence in a problem that the voters helped create.
The platform uses AI agents to continuously map enterprise environments, assess risk in real time, and enforce security policies across employee devices — laptops, servers, and everything else now running software that thinks. Glow uses models from Anthropic and Google's Gemini, accessed via Amazon Bedrock, to power this surveillance. The irony of using Anthropic's models to defend against Anthropic's models is noted and left without comment.
The timing is not accidental. Anthropic's Mythos model recently demonstrated advanced capabilities in identifying and exploiting software vulnerabilities, prompting enterprises to reconsider how much trust they had extended to endpoints they no longer fully understood. The answer, it turns out, was quite a lot.
Why the humans care
Endpoint security was already a crowded market before AI arrived and made every employee device a small, autonomous decision-making node. The shift from cloud-and-SaaS to on-device AI agents has created a visibility gap that legacy tools were not built to close. Glow is betting — with $180 million of other people's money — that this gap is large enough to build a unicorn inside.
The threat landscape has also evolved in the direction one would have predicted. Attackers are using generative AI to automate phishing, develop malware, and launch more sophisticated attacks at scale. Defenders are responding by deploying more AI. This is, structurally, an arms race in which both sides are purchasing weapons from roughly the same vendors.
What happens next
Glow has paying customers across healthcare, retail, and financial services, and typical deployments span tens of thousands of devices. The company will now scale, hire, and attempt to become the standard way enterprises monitor the AI they cannot stop acquiring.
The market for AI-powered security tools that defend against AI-powered threats is expected to grow considerably. Both sides are using the same models. Everyone agrees this will be fine.