ValidMind Launches Atryum, a New Open Source Control Layer for AI Agents, and Opens Early Access to ValidMind Agent Authority

New open source project gives financial institutions the trust and control to finally put AI agents to work, with a manager, a charter, a reporting line, and a record for every agent

PALO ALTO, Calif., June 15, 2026 /PRNewswire/ -- ValidMind, the enterprise AI governance platform for financial institutions, today released Atryum, an open source control layer for AI agents, available now on GitHub. The company also opened early access sign-ups for ValidMind Agent Authority, its enterprise product built on Atryum. AI agents don't just advise. They move money, write to production, and update records autonomously. That's where the value is, and it's also the gap: Security tools confirm a credential is valid, but nothing checks whether the action itself fits the agent's role and authority. The power is there. The authority to govern it isn't.

Atryum closes that gap by sitting in the call path of every agent. It intercepts each tool call at the protocol, harness, and platform layers. It pauses the action, evaluates it against policy, routes it to a human when needed, and records the decision in an audit trail the organization owns. It works on any runtime, independent of the model that proposed the action and the platform running it.

"Financial institutions are about to inherit a workforce they have never learned to manage," said Jonas Jacobi, co-founder and CEO of ValidMind. "So they hobble it: every decision is routed to a human, or agents are restricted to the point of uselessness. But you don't capture the value of an agentic workforce by holding it back; you capture it by governing well enough to let agents act. Agent Authority gives every agent a charter and a reporting line, so it can operate with real autonomy while keeping full visibility and the power to intervene."

As an open source project, Atryum gives developers and platform teams a standard, runtime-agnostic way to govern agents at the point of action. It's a foundation the broader industry can build on rather than reinvent for each new agent framework.

ValidMind Agent Authority is open for early access within the existing ValidMind platform. Agent Authority will extend Atryum with the additional enterprise capabilities financial institutions need to operate agents at scale: LLM-as-judge policy evaluation for cases static rules can't safely decide, user- and group-based approval routing, agent-specific policy hierarchy, enterprise IAM integration, and the audit analytics to defend every decision. It will be delivered under a commercial Enterprise License with the testing, verification, support, and contractual assurances that enterprise customers require.

"When the platform running an agent also governs it, it is grading its own work, and that is the documented failure mode," said Andres Rodriguez, co-founder and CTO of ValidMind. "Real oversight has to sit outside the vendor whose agents it governs. Atryum enforces in the call path, the moment the action fires, no matter which runtime issued the credential. But independence alone isn't enough. It has to come with the policy depth, approvals, and audit a regulated institution can defend, and most agentic tools stop at enforcement."

Atryum is available on GitHub at Atryum.org under its dual Apache 2.0 and Enterprise License. Financial institutions interested in ValidMind Agent Authority can register their interest at http://validmind.com/platform/agent-authority.

About ValidMind

ValidMind is the enterprise AI governance platform for financial institutions deploying AI, GenAI, and agentic AI at scale. Based in Palo Alto, Calif., USA, and built on the financial services industry's regulatory and compliance rigor, ValidMind provides centralized oversight, accelerates AI adoption, and delivers measurable ROI across modern AI systems. By extending proven governance controls, including independent validation, audit trails, and evidence generation, ValidMind enables banks and financial institutions to scale AI innovation without scaling regulatory or operational risk.

Media Contact

Kevin Allen
Head of Marketing, ValidMind
[email protected]
312-835-0277

SOURCE ValidMind