AI Security

Ent made the Cyber 60 by asking the question EDR can’t: why is this person doing that? The answer cost $100M at seed.

EDR can see that Zoom is running. It can’t see that someone just handed remote control of a laptop to a stranger. Ent, founded in 2025 by the RiskIQ team, runs small AI models on the endpoint to judge intent and step in before the click lands. The bar for a $100 million seed is very high.

Nov 13, 2026 · 4 min read

The list. Ent made the 2026–2027 Fortune Cyber 60 in the early-stage group. Fortune’s profile puts it in San Francisco with CEO Elias “Lou” Manousos, founded in 2025. Fortune reports $100 million in funding. That is the seed round Ent announced when it came out of stealth on June 16, 2026, led by Decibel, with Sequoia, Crosspoint Capital, Craft Ventures, Shield Capital, Felicis and In-Q-Tel joining. The Wall Street Journal said Ent had about 100 people at the time. Valuation and revenue are Undisclosed. Lightspeed, which helps run the Cyber 60, is not listed as an investor.

The founders matter here. Manousos and CTO Brandon Dixon co-founded RiskIQ, which Microsoft bought, and then worked on Microsoft Security Copilot. Investors were not paying for a slide. They were paying for a team that has sold to big security buyers before.

The problem in plain words. Endpoint detection and response, or EDR, was built to catch malware: bad code poking at the operating system. Today’s attacker often skips malware. They use the same tools employees use. Dixon put it simply in Decibel’s founder Q&A: an EDR can tell you Zoom is running, but not that a user just gave remote control of their machine to someone outside the company, or why. Each step looks like normal work. The danger is in the sequence and the intent.

The one technical thing. Ent runs small, specialized AI models on the laptop itself. No round trip to the cloud. Ent says that lets it decide in under a second, before the action finishes. Decibel says the models include local computer vision, so the agent understands what the user actually sees on screen, not just process names. Ent then builds what Manousos calls a “work model” of each role, borrowed from self-driving cars: learn how normal work looks, and flag what breaks the pattern, instead of writing a rule for every case.

The research side is concrete. Ent scientist Ruslan Vaulin presented work at the CAMLIS conference in October on small embedding models that detect bad behavior by meaning. A defender writes, in plain language, what they want to stop, from ClickFix-style lures to someone dumping files where they shouldn’t. The model compares live actions to that meaning rather than to a fixed signature. Ent also says it runs in the customer’s own cloud, so the data stays put, and works next to existing EDR, SIEM and identity tools rather than replacing them.

Why it is winning now. AI agents are now acting on the endpoint too. Ent’s own Cyber 60 release cites the Lightspeed Cyber60 Survey: agents are in production at 77% of surveyed companies, but only 20% of those using or exploring agents say they have a complete inventory. Ent pitches one control point that watches both people and agents. It says it is deployed at Global 2000 customers in hospitality, financial services and defense, and it just hired Kevin Arsenault as chief revenue officer. Those customer claims come from the company. None is named.

The CyberMerge view. Here is the uncomfortable part. Ent is not alone. This series has already covered Glow, which raised big before revenue on prevention-first endpoint, and Bold Security, which runs AI on the device to stop data loss. Island and the enterprise browsers want the same moment from the browser side. CrowdStrike and Microsoft own the agent already sitting on most laptops. “Prevention is back” is now a crowded slogan.

So the $100 million seed is a bet on two hard things. First, false positives. A tool that pops up “are you sure?” too often gets turned off by Friday. Ent’s answer is polite, just-in-time nudges instead of alert floods. That has to hold up across tens of thousands of users. Second, the second agent problem. Every CISO already fights over one endpoint agent. Ent has to win a slot next to the EDR, not instead of it, and prove the extra agent pays for itself.

Bottom line. Ent is asking the right question. “What happened?” is EDR’s job. “Why is this happening, and should it finish?” is the job nobody owns yet. Watch for a named customer, a published false-positive rate, and whether the big EDR vendors ship intent features of their own. If they do, Ent becomes a feature. If they can’t, Ent becomes a category.

Sources