ValidMind helps banks, insurers and other regulated organisations govern their models and AI agents. Its platform tests, documents, validates and monitors statistical, AI and generative AI models and controls what AI agents are allowed to do at the moment they act, so model risk and compliance teams can show regulators a single, audit-ready record.
It is built for the model risk management (MRM) practices of financial services and runs as SaaS, on-premises or customer-managed.
At A Glance
| Vendor | ValidMind |
|---|---|
| Category | AI governance and compliance |
| Identities covered | AI agents with governed identities and charters, plus AI and traditional models |
| Deployment | SaaS, on-premises or customer-managed |
| Agentic AI | Agent Authority runtime enforcement (early access), Atryum open source, MCP governance server |
| Customers | General Bank of Canada, Experian |
| Founded | 2022 |
| Market classification | AI Governance & Compliance |
What ValidMind Does
Traditional model risk management assumes a person reviews each model periodically. AI agents act on their own, so periodic review is not enough; they need policy enforced in real time. ValidMind applies one approach to both, governing agents the same way it governs models.
The result is one record of how each model was built, tested and approved and of every action each agent took, which produces the evidence for supervisory frameworks including SR 26-2, the EU AI Act, SS1/23 and OSFI E-23.
Products
ValidMind Library
Model developers use the open-source Library inside their own development environment to test their work. It runs stress testing, backtesting and scenario analysis, and uses language models to write test descriptions and summaries that stay traceable for audit. It handles statistical, AI and generative AI models, foundation model providers and models a customer brings itself.
Governance, Validation And Monitoring Platform
The platform is where validators, risk managers and auditors do their work, replacing spreadsheets and email with one system of record:
- AI-assisted model documentation
- Enterprise validation workflows, approvals and risk tiering
- Model inventories with audit trails and Responsible AI controls
- Monitoring reports and threshold alerts for production models
- Role-based access aligned to the enterprise operating model
Agent Authority (early Access)
Agent Authority answers a question identity systems leave open: an IAM system can establish who an agent is, but not whether a given business action is allowed. Each agent gets a governed identity, a manager, a reporting line and a charter that defines what it may do, and that authority narrows as it cascades to sub-agents and tools. Each action is allowed, paused or denied; paused actions go to an approver together with the agent’s reasoning, and every request, verdict and approver is logged to an audit trail the customer controls. Policies are written as code and applied to every agent action as it happens. Agent Authority is built on Atryum, ValidMind’s open source runtime control plane for AI agents, licensed under Apache 2.0.
How It Works
ValidMind runs in any cloud or combination of services and connects to the tools a customer already has rather than replacing them.
| Component | What it does |
|---|---|
| Library | Runs tests in the developer’s environment and sends results and documentation to the platform. |
| Platform | Holds inventories, workflows, approvals, risk tiers and audit trails. |
| Agent Authority | Sits in the call path outside the agent, intercepting each tool call, checking it against policy, deciding and recording the outcome. Because it is independent of the model and the agent platform, an agent cannot skip or alter the check. |
| MCP server | Agents built with frameworks such as LangChain and AutoGen connect to ValidMind’s MCP governance server to read governance context and policy while they work. |
Identities And Systems Covered
| AI agents | Governed identity, charter, manager and reporting line for each agent, including sub-agents and tools |
|---|---|
| Models | Statistical, traditional, AI and generative AI models, including third-party models |
| Agent platforms | AWS Bedrock Agents, LangChain, AutoGen and custom multi-agent frameworks |
| Data and ML platforms | AWS SageMaker, Databricks and Snowflake |
Company Background
| Founded | March 2022, by Jonas Jacobi, Andres Rodriguez and Mehdi Esmail |
|---|---|
| Headquarters | Palo Alto, California, United States |
| Leadership | Jonas Jacobi, Co-founder and CEO; Andres Rodriguez, Co-founder and CTO |
| Partners | AWS, Experian (Ascend Platform) and Genpact |
| Recognition | Included in Chartis RiskTech100 (2026) and Chartis RiskTech AI 50; an award at the Risk Technology Awards |
Note: This overview is compiled from ValidMind’s publicly available information as of 3 November 2026. It is provided for information only. NHI Mgmt Group does not endorse any vendor, and nothing on this page represents the opinions of NHI Mgmt Group.