The round. Sentra raised $150 million in a Series C led by Capital One Ventures, per the company and SecurityWeek. That brings total funding to more than $260 million. Bessemer, Zeev Ventures, Standard Investments, Munich Re Ventures and Key1 Capital came back in. A Capital One representative joins the board. Sentra says it grew more than 300% year over year. It did not disclose a valuation, and neither did anyone else. So I won’t guess one.
Why the lead investor matters. Most venture leads are funds. This one is the venture arm of a large bank that has to explain its data controls to regulators for a living. When a buyer-shaped investor leads, it is usually a vote on the approach, not just the growth chart. That’s the part worth reading closely.
The approach, in plain words. There are two ways to stop an AI agent from leaking data. One is to sit in the middle and inspect every request the agent makes, in real time. The other is to make sure the agent can’t reach the wrong data in the first place. Sentra is betting on the second. It finds and labels sensitive data across cloud, SaaS, warehouses, file shares and the knowledge stores AI pulls from. It maps who can reach it, people, service accounts and agents, including access they inherited by accident. Then it trims that access and deletes stale data nobody should find.
Why that’s the sharper bet. Inline inspection sounds great on a slide. In practice it means sitting in the path of every agent call, across every model and every tool. That is a latency problem, a coverage problem and an outage risk. Data at rest doesn’t move. If the over-shared folder is gone, no agent can leak it, no matter which model or framework it runs on. It’s boring hygiene. Boring hygiene is what actually shrinks blast radius.
Where I’m skeptical. This is the old DSPM pitch with a new word in front of it. DSPM means data security posture management: find the sensitive data, show who can reach it. Plenty of vendors do that. The category also has a giant in it. Cyera says it raised $600 million at a $12 billion valuation, and Globes reported talks at about $20 billion. Sentra has to win against that balance sheet, plus every platform adding a data tab. Growth of 300% is strong. It’s also easier from a smaller base, and Sentra didn’t say how big that base is.
The real test. Finding data is a feature. Fixing access is the product. Plenty of tools produce a scary report of a million over-shared files and then hand it to a team that will never get through it. Sentra says it revokes risky access and removes sharing automatically, and routes the hard cases to data owners. If that works at scale, it’s valuable. If it needs a human for every change, it’s a dashboard.
For buyers. Ask three things in a pilot. How many risky permissions did it actually remove in 30 days, not just find? Does scanning run inside your environment, so your data never leaves? Sentra says it does, and only metadata goes out. And can it show, after an incident, exactly what an agent could have reached? That last answer is what your lawyers will want first.
Where I land. The money in AI security is splitting into two camps: watch the agent, or shrink what the agent can touch. Both will exist. But the second one keeps working when the agent stack changes next quarter. A bank leading this round says the people with the most to lose agree.