The fast growth of advanced artificial intelligence (AI) models has caused a complicated mix of national security worries, global competition, and legal arguments about how to control these technologies.

進階人工智慧 (AI) 模型的快速增長，導致國家安全憂慮、全球競爭以及關於如何監管這些技術的法律爭論交織在一起，情況非常複雜。

There is currently a strong tension between what AI can do and how it is regulated. For example, Anthropic's 'Mythos' and 'Fable 5' models showed an amazing ability to find weaknesses in secure government networks very quickly. Consequently, the Trump administration introduced strict export controls to stop foreign citizens from using these models. This has caused problems for several organizations; the National Security Agency lost access to important tools, and a legal tech company called Legion has filed a lawsuit, claiming that the sudden loss of these models has seriously damaged its business.

目前 AI 的能力與監管之間存在強烈的緊張關係。例如，Anthropic 的 「Mythos」與 「Fable 5」模型展現了驚人的能力，能快速發現政府安全網路的漏洞。因此，川普政府引入了嚴格的出口管制，以阻止外國公民使用這些模型。這導致了數個機構面臨問題；國家安全局失去了重要工具的使用權，而一家名為 Legion 的法律科技公司則提起訴訟，聲稱突然失去這些模型嚴重損害了其業務。

At the same time, competition between the United States and China is increasing. Anthropic has accused Alibaba of using a 'distillation attack' to steal AI capabilities. On the other hand, Chinese leader Li Qiang warned that if there is no global agreement on governance, the world could lose control over this technology. Despite these conflicts, some experts suggest that the U.S. and China should work together. They argue that risks, such as automated cyberattacks and the creation of biological weapons, require a shared safety plan similar to nuclear agreements during the Cold War.

與此同時，美國與中國之間的競爭正在加劇。Anthropic 指控阿里巴巴使用 「蒸餾攻擊」 (distillation attack) 盜取 AI 能力。另一方面，中國領導人李強警告，若缺乏全球性的治理協議，世界可能會失去對這項技術的控制。儘管存在這些衝突，部分專家建議美國與中國應共同合作。他們認為，如自動化網路攻擊和生物武器開發等風險，需要一個類似冷戰時期核協議的共同安全計劃。

Furthermore, intelligence agencies from the 'Five Eyes' alliance and Germany have noticed a major change in cyber risks. Because AI can now find software vulnerabilities much faster, protecting systems has become a top priority for leaders rather than just a technical task. While some industry leaders argue that companies often focus on technical success before thinking about ethics, most international security agencies agree that the time to create effective safety rules is running out quickly.

此外，「五眼聯盟」的情報機構與德國已注意到網路風險的重大變化。由於 AI 現在能更快地發現軟體漏洞，保護系統已成為領導者的首要任務，而非僅僅是一項技術工作。雖然部分產業領袖認為，公司往往在追求技術成功後才考慮倫理，但大多數國際安全機構均認同，制定有效安全規則的時間已所剩無幾。

The world is currently in a delicate balance, trying to develop powerful AI while avoiding global instability and political conflict.

世界目前處於一種微妙的平衡，試圖在開發強大 AI 的同時，避免全球不穩定與政治衝突。