HL7 International embraces artificial intelligence as a powerful enabling technology for health data interoperability. Rather than restricting AI use of our standards, we are actively engineering our processes and content to be AI-ready. We believe that AI systems capable of understanding FHIR® and other HL7 standards ultimately serve our mission to improve health and well-being through better data exchange.
HL7 International's mission is to create and promote the adoption of innovative interoperability standards that improve health and well-being while fostering a diverse and inclusive global community. We pursue this so that people everywhere can live in optimal health. Because open standards enable new digital freedoms, we want our standards to be as widely used and as deeply understood as possible. By humans and machines alike. Standards demonstrate the network effect — they become more valuable the more widely they are adopted.
Some standards development organizations (SDOs) restrict or prohibit the use of their intellectual property in AI training and inference. HL7 International takes a different approach. We view AI systems as legitimate and valuable consumers of our standards — partners in the global project of interoperability rather than threats to our business model.
Our licensing choices reflect this view. HL7 publishes our FHIR platform standard and related specifications under the Creative Commons Zero (CC0) public domain dedication, the most permissive licensing approach available. CC0 reflects a deliberate strategic commitment: by removing all licensing barriers, we maximize adoption potential and minimize friction for all implementers, including AI developers and the AI systems they build.
HL7's flagship FHIR standard is unique not only as a modern API standard for healthcare, with innovations in open standards development, but also in its structure and publication format. FHIR reflects a fundamentally developer-native approach, setting it apart from most other health IT standards. Rather than static, paywalled PDFs, FHIR is published as a fully navigable website where every resource, data type, and operation has its own structured page — paired with computable, machine-readable definitions in JSON and XML that allow tools to programmatically interrogate the standard itself. Further, the platform specification and derivative implementation guides, along with their computable parts, are distributed as versioned NPM packages via a public registry, enabling reproducible builds and automated dependency resolution familiar to any software developer.
Additionally, we have embraced the expectation of co-developing and testing an open-source (typically licensed under Apache 2.0) reference implementation software alongside the standard. This ensures that the published standard is actually implementable, and gives implementers working software to learn from, test their code against, or even incorporate into their products.
Because FHIR is freely available, extensively documented online, and has corresponding open source reference software code bases on public platforms like Github, FHIR is already embedded in the training corpora of the world's leading large language models. Developers today can ask an AI assistant to generate a FHIR patient resource, write a FHIR search query, or explain the semantics of a SMART on FHIR authorization flow — and receive accurate, useful answers. This is a natural consequence of our open-first standards strategy at work.
In short, we see the standard not as a document to be read, but as an open platform to be built upon.
As powerful as FHIR's publication format is, we've also found that its structural verbosity (complex HTML pages, JSON/XML schemas) makes it suboptimal for consumption by many current Large Language Models (LLMs). We want to have standards developers and implementers aided by the best available tools. So, earlier this year, we began publishing experimental LLM-friendly representations of our specifications, including the specification drafts that are up for ballot (public comment).
HL7 now provides AI-optimized package bundles formatted in Markdown — reducing noise, improving token efficiency, and making it substantially easier for AI systems to ingest, index, and reason about the full content of the standard. These bundles are available publicly. For example, the FHIR US Core AI package is available at: https://hl7.org/fhir/us/core/ai.zip.
When HL7 International freely licensed its standards over a decade ago, we transitioned from a business model in which membership value was derived from access to standards to one in which membership unlocks a global community of innovators and a world-class governance system. We have public domain licensing, but protect our intellectual property through trademark rather than copyright restriction. The FHIR® mark is a registered trademark; use of the FHIR brand to describe products and services is governed by HL7's trademark policy to ensure quality and prevent misrepresentation. This approach protects the integrity of the standard without impeding access to it.
The open, AI-friendly availability of HL7 standards is already catalyzing AI development. For example, Anthropic created a Claude "Agent Skill" for FHIR-based development, helping developers build digital health innovations with AI assistance grounded in HL7 standards. Many similar examples are emerging. This kind of ecosystem development, where AI toolmakers invest specifically in standards competency, is only possible when the standards themselves are freely accessible.
HL7 views the emergence of AI-powered FHIR development tools as a validation of our open-standards strategy and a direct amplifier of standards adoption. When AI coding assistants understand FHIR, every developer who uses those assistants becomes a more capable FHIR implementer.
HL7 International urges the standards community and policymakers to carefully consider the tradeoffs between IP protection and public good infrastructure. Prohibiting AI use of standards might protect the content, but it ultimately diminishes their relevance. In a world increasingly shaped by AI-assisted development, standards that succeed will be those AI systems know best. HL7 International is committed to ensuring that our standards are as widely used and as deeply understood as possible — by humans and machines alike.