Skip to content

OpenAI Advocates for AI System Evaluation by Civil Authorities Instead of NSA

In a nutshell: OpenAI calls for mandatory evaluations of AI models by the Commerce Department rather than intelligence agencies, leveraging established relationships with leading labs.

In a new policy paper, OpenAI disputes the recently signed White House Executive Order on two key points and urges Washington to strengthen the Center for AI Standards and Innovation (CAISI) instead of the National Security Agency in overseeing advanced AI systems.

The White House issued a voluntary assessment directive on Tuesday for advanced AI systems’ cybersecurity risks, led by the National Security Agency. OpenAI instead proposes that mandatory evaluations of advanced AI models take place, but be conducted by the Center for AI Standards and Innovation (CAISI) at the National Institute of Standards and Technology within the Commerce Department.

The company argues that CAISI has already established working relationships with leading AI companies: OpenAI, Anthropic, and other labs have already committed to sharing information about new models with CAISI. The NSA does not have such established connections. OpenAI spokesperson Chris Lehane further raised concerns about the White House’s plan to develop assessment criteria on a classified basis, as this could leave companies unclear about when their models fall under NSA oversight.

CEO Sam Altman met with White House officials and congressional members from both parties on Wednesday to present OpenAI’s positions. Lehane suggested that the next 60 days—the timeframe the Executive Order gives agencies to develop details—could be critical for incorporating OpenAI’s approaches into the final regulation. Congressional members are also considering plans to expand CAISI’s authority.

OpenAI further advocates for mandatory evaluation processes in which no single lab can unilaterally decide, but currently rejects more aggressive measures—such as government approval requirements before model release. Lehane justified this by saying that a robust evaluation framework must first be established before more extensive regulations can be discussed.


Source: www.politico.com · Published June 3, 2026
Lumi AI News — AI-assisted curation according to Article 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.2.9.

Share on: