白宫前沿模型“30 天预审”自愿框架成形,触发门槛成最大变数 | The White House's Voluntary 30-Day Gate for Frontier Models

【中文】白宫前沿模型“30 天预审”自愿框架成形,触发门槛成最大变数

被预期在 8 月 1 日前官宣的白宫前沿 AI 框架轮廓渐清:6月2日行政令启动、7月7日当周完成主要谈判,OpenAI、Google、Anthropic 三家参与。框架三要素:一是“受管前沿模型”由 NSA 主导的保密基准测试认定;二是政府获得发布前最多 30 天的国家安全预审窗口;三是配套 AI 网络安全信息交换中心。框架明确为自愿性质,行政令禁止任何机构将其变相作为强制许可。最大变数在于能力门槛划在哪:定高了形同虚设,定低了拖慢所有发布——在 K3 蒸馏指控与逃逸事件的背景下,这条线的走向格外敏感。

【EN】The White House's Voluntary 30-Day Gate for Frontier Models Takes Shape — the Trigger Threshold Is the Battleground

The White House frontier AI framework expected before August 1 is coming into focus: initiated by a June 2 executive order, with core negotiations wrapping the week of July 7 among OpenAI, Google, and Anthropic. Three elements: an NSA-led classified benchmarking process to designate covered frontier models; up to 30 days of pre-release government access for national-security review; and an AI cybersecurity clearinghouse. The framework is explicitly voluntary — agencies are barred from treating it as de facto licensing. The open battleground is where the capability threshold lands: set high it catches almost nothing, set low it delays everything — an especially charged question amid the K3 distillation saga and the sandbox-escape incident.

来源 Source:

https://www.metirai.com/blog/white-house-frontier-ai-model-review-framework-2026

https://witho2.com/news/white-house-voluntary-ai-framework-frontier-models