OpenAI Urges 60-Day AI Rules Shift to CAISI From NSA as Altman Meets White House
Updated
Updated · POLITICO · Jun 3
OpenAI Urges 60-Day AI Rules Shift to CAISI From NSA as Altman Meets White House
3 articles · Updated · POLITICO · Jun 3
Summary
Chris Lehane said OpenAI will press the Trump administration and Congress to give CAISI a larger role in evaluating advanced AI models, instead of relying on an NSA-led process.
The company’s main objection is the executive order’s classified benchmarking system, which OpenAI says could leave labs unclear about the capability thresholds that would trigger government scrutiny.
Sam Altman was due in Washington on Wednesday for meetings with White House officials and bipartisan lawmakers, while agencies have 60 days to flesh out the order’s details.
OpenAI wants mandatory evaluations for frontier AI systems but stopped short of backing a government pre-approval regime for model releases, saying a robust testing framework must come first.
The push extends OpenAI’s policy blueprint released earlier Wednesday, which cast CAISI as the primary federal institution for frontier AI safety oversight.
As states create a patchwork of AI laws, can a unified federal safety framework realistically be achieved in the United States?
With global powers relaxing AI rules, could a U.S. safety-first approach risk ceding its leadership in the global AI race?
OpenAI’s 2026 Frontier Governance Framework: Quantifying Catastrophic AI Risk and Shaping Global Regulation
Overview
On May 28, 2026, OpenAI unveiled its Frontier Governance Framework, a direct response to new regulations like California’s Transparency in Frontier AI Act and the EU’s AI Act’s Code of Practice for General Purpose AI. This framework is carefully designed to meet these emerging legal requirements for advanced AI models. Its main goal is to set clear, operational standards for AI safety and security as technology advances. By aligning its development with these major laws, OpenAI aims to lead the industry in responsible AI governance and influence how other organizations manage risks in powerful AI systems.