OpenAI has published a federal governance blueprint proposing mandatory pre-release evaluations of frontier AI models by the Center for AI Standards and Innovation (CAISI), while insisting that final deployment authority should remain with developers rather than regulators. The proposal calls for annual third-party audits, public transparency reports, mandatory incident reporting, cybersecurity protections for model weights, and whistleblower safeguards. It also recommends that federal agencies be prohibited from procuring products built on unevaluated frontier models. Analysts note the framework could entrench large developers by letting them help define the evaluation thresholds others must meet, and that the proposal functions more as an evidence-producing regime than a true accountability mechanism.
