Beyond the model: A data foundation for a trustworthy AI future in the US

A central tenet of the AI Action Plan is the creation of a national "AI Evaluations Ecosystem." Public trust in AI hinges on our collective ability to verify the outputs of systems that can influence many aspects of our lives. The plan's mandate to identify and mitigate ideological bias, ensure objectivity, and test for robustness is a direct call for transparency across systems.

Implementation of this policy requires the ability to ask deep questions of our AI systems and their data. To do this, government and private sector partners need a powerful search and analytics engine capable of providing deep, real-time visibility into these complex black boxes. Before using their proprietary data with AI, agencies must be able to search, access, and analyze their data holistically, regardless of what format it’s in or where it’s stored. Then, once AI outputs have been delivered with the data, agencies must be able to track performance metrics, trace data lineage, and instantly search petabytes of model-generated content to identify the patterns, anomalies, and vulnerabilities that could erode public trust. This is how the abstract goal of trustworthy AI becomes a concrete, achievable reality.

The AI plan also outlines the urgent need for a secure and resilient national infrastructure of data centers and energy resources, demonstrating that AI is often a dual-use technology. This new national infrastructure will not only be a target for cyber attacks but also for novel, AI-specific threats like data poisoning and model evasion, potentially executed by the same nation-state actors that are employing traditional cyber attacks today.

A resilient national AI infrastructure requires greater visibility. The only way to defend against unknown threats is to unify security data from across the entire AI technology stack from the cloud environment to the network to the endpoint. A security solution built on a unified data platform provides this comprehensive view, allowing public-private defense teams to detect threats and respond at speed. This commitment to data visibility and interoperability is the foundation of a secure-by-design approach.

The AI Action Plan creates a fork in the road for government technology policy. One path leads to reliance on a few large, closed AI systems that create risks of vendor lock-in and government waste. The other path, highlighted by the plan’s support for open source principles, leads to a more resilient and competitive environment built on interoperable components.

The US government's long-term interests are best served by choosing the second path. An open, standards-based approach ensures agencies can select the best tools for their mission without being tied to a single provider. This provides better value for the taxpayer and, most importantly, enhances national resilience by creating a more diverse and adaptable technology supply chain. By providing a common data platform that can connect to any model — open or proprietary — Elastic can help the government maintain control over its data and build an AI future that is both innovative and secure.

The race to AI leadership will not be won by simply acquiring the most powerful models. It will be won by the nation that builds the most transparent, manageable, and secure data foundation. That is the challenge of our time, and Elastic is committed to helping America meet it.