OpenAI calls for US to lead global AI safety standards

By Splitfeed EditorialLast updated Sep 26, 2026, 20:18 UTC

On September 21, 2026, OpenAI published a position paper calling on the United States government to lead a coordinated international effort to set shared AI safety standards. The proposal covers three main areas: evaluation frameworks for frontier AI models, reporting requirements for developers, and governance structures, with explicit attention to systems capable of recursive self-improvement. On the same day, OpenAI and the General Services Administration announced an extension of discounted AI services to local governments, a move that adds a commercial dimension to what was otherwise framed as a policy initiative.

The announcement carries weight because it arrives at a moment when OpenAI is navigating what at least one outlet characterized as a safety backlash directed at the company. A position paper urging governments to build oversight structures is a notable posture for a company that has faced criticism over whether its own internal safety commitments are credible. The core dispute is whether a proposal like this represents genuine advocacy for accountability or a strategic attempt to shape regulation in ways that favor established frontier developers, who have more resources to meet complex compliance requirements than smaller competitors or open-source projects.

The Pro-AI camp has not published reactions to this story yet. Typically, advocates in that camp would welcome the US taking a leadership role on global standards, arguing that coordinated frameworks reduce fragmentation and create clearer rules for beneficial development. The Anti-AI camp has also not reacted publicly yet, though critics in that camp tend to treat company-authored safety proposals with skepticism, questioning whether they amount to self-regulation dressed in policy language. From the Middle Ground camp, Techmeme flagged the paper's position that fully autonomous recursive self-improvement is not something being pursued today and should not be unless it can be done safely, suggesting the proposal's most significant technical commitments may be in what it rules out rather than what it endorses.

The argument will sharpen if and when any US government body responds formally to the proposal or signals whether it intends to use the framework as a starting point for legislation or international negotiation. The GSA partnership with local governments could also develop into a test case for how OpenAI's commercial reach interacts with its stated governance goals.