政府要求在安全漏洞發生後提前獲取前沿人工智慧模型權限
Government Demands Early Access to Frontier AI Models After Security Breaches
Updated at: August 8, 2026 at 01:30 AM
2026年6月2日，美國政府在科技政策上採取了重要一步，發布了「促進先進人工智慧創新與安全」行政命令。
On June 2, 2026, the U.S. government took a significant step in tech policy by introducing the 'Promoting Advanced Artificial Intelligence Innovation and Security' executive order.
此政策為政府建立了一個機制，以便在最強大或稱之為「前沿」AI模型發布前取得存取權。
This policy establishes a framework for the government to gain pre-release access to the most powerful, or 'frontier,' AI models.
雖然先進AI能增強國家網路防禦，但其利用軟體漏洞的能力也構成了嚴重安全風險。
While advanced AI can strengthen national cyber defenses, its ability to exploit software vulnerabilities also poses a serious security risk.
為緩解此風險，大型開發商現在可以自願在模型公開發布前，向聯邦機構提供長達30天的存取權限。
To mitigate this, major developers can now voluntarily provide federal agencies access to their models for up to 30 days before they are released to the public.
雖然參與形式上是自願的，但專家認為這對頂尖企業而言，實際上是一項必要要求。
While participation is technically voluntary, experts suggest it functions as a practical requirement for top-tier companies.
為確保過程安全，政府使用一套機密基準測試系統來決定哪些模型需要審查。
To keep the process secure, the government uses a classified benchmarking system to determine which models need review.
此舉作為「金鷹」計畫等更廣泛項目的一部分，代表政府在日益複雜的數位環境中，努力平衡快速AI創新與必要的國家安全保障措施。
This move, part of broader projects like the 'GOLD EAGLE' initiative, represents an ongoing effort by the administration to balance rapid AI innovation with necessary national security safeguards in an increasingly complex digital landscape.
