1. News
  2. AI models & agents
  3. Sentra · AI security
Sentra · AI security3 min read

Sentra raises $150M to control what AI agents can see in company data

Data security startup Sentra raised $150 million to limit what AI agents can reach in company files, a risk for any business letting AI tools read its data.

By the Dashline deskSource: Techmeme
Sentra raises $150M to control what AI agents can see in company data
Photo: Pauli Rautakorpi, CC BY 2.0

Sentra, a data security startup, has raised $150 million to build software that controls what AI agents can see inside a company's data. Capital One Ventures led the round, which is a Series C. Sentra says investors have put more than $250 million into the company since it launched in 2021.

The problem it targets

An AI agent is a program that works on its own, such as reading files to complete a task. Sentra says every agent a company turns on is another identity reading its data, often around the clock, with no central control over it. Sentra's chief executive, Yoav Regev, called agents "the busiest consumers of enterprise data." He says many inherit access to data that was never fully sorted or governed.

How the software works

Sentra builds a map of which people, service accounts and AI agents can reach a given set of data, including access passed down from other settings. A service account is a login used by software instead of a person. The company then acts in several steps:

  • It removes stale and high-risk data that no agent has any reason to see.
  • It gives service accounts and agents only the minimum access their work needs.
  • It flags an agent that reaches for sensitive data it does not normally touch.
  • It can strip the permissions or sharing involved on its own, and sends the worst cases to the data owner.

After an incident, legal and compliance teams get a record of everything the agent could have reached.

Where the data stays

Sentra says its analysis runs inside the customer's own environment, so sensitive data never leaves it. The company argues that tools which copy data into a vendor's systems create a second pile of data to protect. It says that adds risks around how long data is kept and where it is stored, especially in regulated industries.

The money and the claims

Capital One Ventures will take a seat on Sentra's board. Existing investors Bessemer Venture Partners, Zeev Ventures, Standard Investments, Munich Re Ventures and Key1 Capital also joined. Sentra did not disclose a valuation. The company says its business grew more than 300% over the past year, and that only a small minority of organizations have controls over what their agents touch. Both figures come from Sentra itself.

The new money will go toward better visibility into the agents, copilots and models touching sensitive data, and faster containment when an agent behaves unexpectedly. Sentra also plans to grow its field and channel programs and expand into new regions.

What to do

If your business lets AI tools read company files, check what they can reach first. Many tools inherit far more access than their job requires. Start by asking who or what can open your most sensitive folders, and cut back anything that does not need to be there.

The takeaway

Before letting an AI tool read company files, check what it can reach and trim its access to only what the job needs.

More articles

You should really read these.

Next step

Let’s talk