top of page

Virtru Launches Data Harbor to Enforce AI Agent Access at the Data Layer

13 minutes ago
1 min read

Virtru has launched its Data Harbor Platform, a data-centric security system intended to restrict what people and AI agents can open even after they have access to an application. The company unveiled the platform in Washington, D.C., on November 17.


Built on the open Trusted Data Format standard, Data Harbor binds encryption and access policies to sensitive data. Virtru says an agent's identity and entitlements determine whether a protected file can be decrypted, with policy enforcement continuing as the data moves across applications or organizations. The approach is designed to complement controls around agent runtimes, identities and gateways.


Each grant or denial produces a signed, tamper-evident record of the requester, applicable policy and decision, according to the company. Security teams and auditors can use the records to check how access was enforced. Virtru also says organizations can revoke access to protected files after sharing them.


Customers can run the platform in their own environments and integrate its controls through SDKs and APIs. Virtru named Booz Allen, Kong, Mattermost, Pexip, Rocket.Chat and ThunderCat among partners building on the platform at launch.


The announcement addresses a practical challenge for enterprise AI deployments: an agent may have valid credentials to a system while still needing narrower rights to the data inside it. Virtru's architecture places an additional decision at the data tier. Organizations evaluating the platform should test how its policy model, key management, audit evidence and revocation behave in their own workflows.

bottom of page