OpenAI co-founds Appia Foundation hosted by Linux Foundation to build AI assessment and governance standards

OpenAI announced it co-founded the Appia Foundation, hosted by the Linux Foundation, to develop open, modular specifications for evaluating, assessing, and governing increasingly capable AI systems. Appia will translate international standards and established frameworks into practical assessment criteria across the AI value chain, creating a trust layer through which third parties can verify conformity to standards. The effort aims to develop a shared technical language that allows national and international institutions to trust each other's AI governance work.

OpenAI has already put standards-building principles into practice through testing partnerships with US CAISI and UK AISI, publishing a shared playbook for trustworthy third-party evaluations. That playbook sets disclosure standards for frontier capability assessments: the system tested, its tool access, evaluation harness, methods for eliciting capabilities, resources available, and validation checks performed. These practices have led to concrete improvements in OpenAI's systems based on feedback from independent evaluators.

Appia's work complements OpenAI's broader Preparedness Framework, which defines how the company operationalizes managing serious risks from advanced AI systems. The Appia effort also aligns with OpenAI's policy recommendations for a durable US framework, a strengthened Center for AI Standards and Innovation (CAISI), and international cooperation on compatible safety frameworks, trusted channels for sharing risk findings, and coordinated responses to incidents.

For architects and policymakers, the Appia Foundation represents an attempt to reduce fragmentation in AI governance by creating reusable, interoperable assessment criteria and evidence. A shared technical language and trusted evaluation practices can help national institutions evaluate frontier systems independently, then recognize and coordinate on each other's findings. This standardization approach mirrors how aviation safety and financial services established trusted evidence chains across jurisdictions.