OpenAI helped found the Appia Foundation, hosted by the Linux Foundation. Appia aims to develop open, modular specifications that translate international standards and established frameworks into practical assessment criteria across the AI value chain. These specifications are intended to provide a missing trust layer by which independent third parties can check conformity with standards and produce clearer, reusable evidence when models, infrastructure, and applications are developed by different organizations.
Why this matters now
Increasingly capable models can bolster cyber defense, accelerate scientific discovery, and broaden access to expertise. However, they also carry safety and security risks if their capabilities are misunderstood, safeguards are inadequate, or governments lack the information required to respond. Realizing benefits safely and confidently requires institutions with technical and governance capacity to evaluate, secure, and govern these systems.
Appia's role and goals
Appia’s work focuses on creating interoperable, technically grounded practices that can be used across organizations, jurisdictions, and supply chains. By defining open, modular specifications, Appia seeks to establish a shared technical language enabling national and international institutions to trust one another’s assessments and to make third‑party evaluation more practical and reusable.
Institutional proposals and the international dimension
OpenAI’s recent blueprint for democratic governance of frontier AI outlines a roadmap for this effort: a durable U.S. framework, a strengthened Center for AI Standards and Innovation (CAISI), and a broader resilience strategy across government. The blueprint acknowledges that frontier risks are international in scope, and calls for nations to cooperate on compatible safety frameworks, trusted channels for sharing risk findings, and coordinated incident responses.
National capacity and international cooperation should reinforce each other. Strong national institutions such as CAISI can develop technical expertise, evaluate frontier systems, and support an independent assessment ecosystem. A network of capable national institutions can then establish shared methods, recognize trusted evidence, and provide governments with the common technical understanding needed to act together.
Standards, evaluation practices, and prior collaborations
Standards are central to this work and must be grounded in credible evaluation practice and technical rigor. In a shared playbook for trustworthy third‑party evaluations, OpenAI and partners set out what frontier assessments increasingly need to disclose: the system tested, its tool access and evaluation harness, the methods used to elicit capabilities, the resources available, and the checks performed to validate results.
OpenAI has put these principles into practice through testing partnerships with the U.S. Center for AI Standards and Innovation (US CAISI) and the U.K. Artificial Intelligence Safety Institute (UK AISI). Those partnerships, focused on frontier capability assessments and biological‑misuse safeguards, led to concrete improvements in OpenAI’s systems. Such collaborations build a foundation for practices that can be standardized to check performance in a comparable way.
OpenAI's internal frameworks
These external practices complement OpenAI’s broader safety infrastructure. OpenAI’s Preparedness Framework forms the basis for how the company defines and operationalizes its approach to managing the most serious risks from advanced AI systems, including internal practices. The Frontier Governance Framework translates relevant parts of that approach into a public governance document focused on regulatory obligations such as risk assessment, model reporting, security controls, incident response, and incorporation of external expert input. Together, these artifacts help turn broad commitments into operational practices that can be validated and improved.
Broad engagement in standards forums
OpenAI already contributes across a broader ecosystem of standards and pre‑standardization efforts. The company participates in the International Organization for Standardization and International Electrotechnical Commission Joint Technical Committee 1, Subcommittee 42 on Artificial Intelligence; takes part in the National Institute of Standards and Technology‑led Artificial Intelligence Consortium; helped found the Frontier Model Forum and the Linux Foundation’s Agentic Artificial Intelligence Foundation; participates in the Coalition for Secure Artificial Intelligence; serves on the steering committee of the Coalition for Content Provenance and Authenticity; and engages with Internet Engineering Task Force and Fast Identity Online Alliance processes to advance interoperable technical standards.
Conclusion
Appia’s work, and OpenAI’s participation in related standards efforts, form part of a broader attempt to translate lessons from frontier development into open, technically grounded practices that governments, companies, and independent assessors can use across jurisdictions. The initiative aims to create a reliable measurement and assessment layer that supports trust and cooperation among national and international institutions in managing the safety of increasingly capable AI systems.
