Voluntary standards can't really control anyone, right? In the end, it all depends on the White House's mood.

According to the Financial Times, the U.S. government is in in-depth discussions with leading AI companies such as OpenAI, Anthropic, and Google, and plans to release a series of voluntary standards for AI model publishing as early as next week. Insiders revealed that the measure aims to implement the AI executive order signed by Trump last month, establishing a benchmark evaluation process and release timeline for models with "frontier" cybersecurity capabilities. Previously, due to the Trump administration's export controls on Anthropic's latest model and restrictions on OpenAI's releases, the industry was thrown into chaos. The new standards aim to replace the previous ad hoc interventions by clarifying access permissions, classification benchmark evaluations, and review durations.