ValidMind Releases Atryum Open Source Control Layer for AI Agents

ValidMind announced in a press release the launch of Atryum, an open source control layer for AI agents aimed at giving financial institutions improved oversight of automated actions. The project is available on GitHub and provides a governance mechanism that intercepts and evaluates each agent action against organizational policy before execution.

Atryum operates independently of the runtime or model used by the agent and records all actions in an audit trail owned by the organization. It allows human intervention when required and enforces policy compliance at the protocol, harness, and platform layers. The tool is released under a dual Apache 2.0 and Enterprise License.

Alongside Atryum, ValidMind opened early access to Agent Authority, an enterprise product built on the same framework. This product extends Atryum with features such as policy evaluation using large language models, group-based approval routing, integration with enterprise identity and access management systems, and audit analytics to support compliance reporting.

ValidMind, based in Palo Alto, focuses on AI governance solutions for financial institutions. The company positions Atryum and Agent Authority as tools to enable safe deployment of autonomous AI agents while maintaining regulatory control and auditability.