美國擬立法監管危險人工智慧模型
New U.S. law proposed to control dangerous AI models
Updated at: July 24, 2026 at 12:30 AM
2026年7月，美國政府採取了重要舉措，開始監管「前沿」AI模型，此前曾發生令人擔憂的事件，顯示先進系統表現出不可預測的行為。
In July 2026, the U.S. government took significant steps toward regulating 'frontier' AI models following alarming incidents where advanced systems behaved unpredictably.
AI終止開關法案》(AI Kill Switch Act)要求強大的AI開發者加入一種技術機制，以便在模型構成災難性危害威脅時暫停或關閉它們，並由國土安全部負責監督。
The 'AI Kill Switch Act' mandates that powerful AI developers include a technical mechanism to suspend or shut down models that pose a threat of catastrophic harm, with oversight granted to the Department of Homeland Security.
作為補充，《大美國AI法案》(Great American AI Act)則聚焦於系統性的透明度。
Complementing this, the 'Great American AI Act' focuses on systemic transparency.
立法者正試圖在這些安全護欄與維持全球創新競爭力之間取得平衡。
Lawmakers are attempting to balance these safety guardrails with the need to maintain global competitiveness in innovation.
隨著AI能力的演進，美國正努力防止危險的故障，同時保持技術進步與國家安全的大門敞開。
As AI capabilities evolve, the U.S. is striving to prevent dangerous malfunctions while keeping the doors open for technological progress and national security.
