OpenAI and Anthropic Lobby for Standardized Federal AI Reviews

The companies are pushing for a consistent framework that would apply to all major developers, potentially pulling rivals like Meta and Elon Musk’s xAI under the same oversight. Under the current June 2 executive order, federal agencies including the NSA and CISA must define by August 1 which systems qualify as covered frontier models. Developers would then be able to provide government officials with access to these models for up to 30 days of testing, balancing national security concerns with the need for competitive clarity.

While OpenAI and Anthropic seek predictable benchmarks, the broader industry remains divided. A coalition including Nvidia, Microsoft, and Meta recently warned against sweeping restrictions on open-weight models, arguing that overly broad controls could stifle American innovation and weaken national competitiveness against China. These signatories contend that open-source flexibility is essential for research and business, distinct from the closed-model security reviews prioritized by OpenAI and Anthropic. As the August deadline approaches, federal regulators face the challenge of determining whether to focus on specific, measurable capabilities or broad distribution-based conditions.