Virtue AI PolicyGuard enforces AI policies

Virtue AI said April 28 it launched PolicyGuard, a product that turns written enterprise AI policies into runtime controls for models, agents, and applications. (prnewswire.com) The company said customers can define policies in natural language or upload existing PDFs, websites, and JSON files, then enforce those rules without building custom engineering pipelines. (prnewswire.com) (helpnetsecurity.com) Virtue AI said PolicyGuard can stack more than 30 compliance frameworks, including the European Union Artificial Intelligence Act, the General Data Protection Regulation, the Financial Industry Regulatory Authority, and MLCommons. (prnewswire.com) (helpnetsecurity.com) The basic problem is that many companies already have “acceptable use” documents for AI, but those documents do not automatically stop a model or agent from taking a forbidden action in production. Virtue AI said the risk now sits in tool use, application behavior, and multi-step agent workflows, not only in chatbot text. (prnewswire.com) (helpnetsecurity.com) PolicyGuard is built to act as that enforcement layer. Virtue AI said it can evaluate content in its original language, show the reasoning behind decisions, and run on-premises, in the cloud, or as software as a service. (prnewswire.com) (helpnetsecurity.com) Chief executive Bo Li said, “Enterprises already have AI policies. The challenge is enforcement.” He said PolicyGuard can define and enforce organization-specific rules from “a simple PDF upload, or a few lines of natural language.” (prnewswire.com) The release follows Virtue AI’s January 29 launch of AgentSuite, which the company described as a multi-layer security and compliance platform for enterprise AI agents. In that announcement, Virtue AI said 79% of enterprises are already deploying AI agents while 97% lack proper security controls, citing IBM. (prnewswire.com) (virtueai.com) AgentSuite focused on red-team testing, tool validation, runtime alerts, access control, and audit trails for autonomous systems that can modify databases, trigger payments, and reach sensitive systems. PolicyGuard adds a way to translate policy text into the rules those systems are supposed to follow. (prnewswire.com 1) (prnewswire.com 2) The company’s pitch is that AI governance should sit above any single model choice, with one traceable policy layer deciding what agents and applications may do. On April 28, Virtue AI said PolicyGuard is meant to supply that layer before an audit or incident forces a rewrite. (prnewswire.com)