OpenAI and the Appia Foundation: Establishing Shared Standards for Advanced AI

OpenAI has helped found the Appia Foundation, hosted by the Linux Foundation, to create open, modular specifications that translate international standards into practical assessment criteria for AI. This initiative aims to build a critical trust layer by allowing third parties to verify conformity with standards and creating a shared technical language for national and international institutions.

The Role of the Appia Foundation

The Appia Foundation is designed to address the gap between high-level international standards and practical, technical implementation. By developing modular specifications, Appia will provide a framework for which third parties can check conformity with standards, producing reusable evidence when models, infrastructure, and applications are developed by different organizations.

This interoperability is intended to ensure that safety and security risks are managed consistently across the AI value chain, regardless of the jurisdiction or organization involved in development.

Integration with National and International Governance

OpenAI asserts that national capacity and international cooperation must reinforce one another. The organization advocates for a durable U.S. framework and a strengthened Center for AI Standards and Innovation (CAISI), which can develop technical expertise and support an independent assessment ecosystem.

According to OpenAI, a network of capable national institutions can establish shared methods and recognize trusted evidence, providing governments with the common technical understanding necessary for coordinated responses to incidents and compatible safety frameworks.

Technical Rigor in Evaluation Practices

Standards for advanced AI must be grounded in credible evaluation practices. OpenAI has established a shared playbook for trustworthy third-party evaluations, which mandates the disclosure of the following:

• The system being tested
• Tool access and the evaluation harness
• Methods used to elicit capabilities
• The resources available
• Checks performed to validate results

These principles have been applied through testing partnerships with the US CAISI and UK AISI, which focused on frontier capability assessments and biological-misuse safeguards to drive concrete system improvements.

OpenAI's Internal Safety Infrastructure

The Appia Foundation's work complements OpenAI's existing internal safety frameworks:

• Preparedness Framework: The foundation for defining and operationalizing the management of the most serious risks from advanced AI systems.
• Frontier Governance Framework: A public governance document focusing on regulatory obligations, including risk assessment, model reporting, security controls, and incident response.

By combining these internal practices with the Appia Foundation's open specifications, OpenAI aims to translate lessons from frontier development into technically grounded practices that can be used by governments, companies, and independent assessors globally.