The United States government has ordered restrictions on foreign access to Anthropic's Fable 5 and Mythos 5 AI models. As a result, these systems have been suspended globally.

美國政府已下令限制外國訪問 Anthropic 的 Fable 5 與 Mythos 5 AI 模型。因此，這些系統在全球範圍內已被暫停。

The restrictions began on June 12 after the U.S. Commerce Department issued an export control order. The government emphasized national security concerns, specifically the risk of 'jailbreaking,' where users use special prompts to bypass safety rules. This action was caused by reports that a South Korean telecommunications company with suspected links to China had access to Mythos 5, as well as safety warnings from Amazon. Although Anthropic claimed these issues were minor and already fixed, the company stopped all global access to follow the law, which also affected its own foreign employees.

限制始於 6 月 12 日美國商務部發布出口管制令後。政府強調國家安全考量，特別是「越獄」風險，即使用者利用特殊提示詞繞過安全規則。此舉是由於有報告指出，一家涉嫌與中國有聯繫的韓國電信公司可訪問 Mythos 5，以及亞馬遜發出的安全警告。儘管 Anthropic 聲稱這些問題微不足道且已修復，但公司為遵守法律仍停止了所有全球訪問，這也影響了其外籍員工。

This decision has caused tension between the government and the tech industry. More than 170 executives asserted that these rules could weaken U.S. leadership in technology by taking away essential tools from cybersecurity experts. Furthermore, this happened just as Anthropic was preparing to go public (IPO), which could affect investors despite the company's high estimated value. The relationship between the company and the state has already been difficult due to past arguments with the Department of Defense over mass surveillance and autonomous weapons.

這一決定引起了政府與科技產業之間的緊張關係。超過 170 位高階主管主張，這些規則可能會奪走網絡安全專家的核心工具，從而削弱美國在科技領域的領導地位。此外，此事發生在 Anthropic 準備首次公開募股（IPO）之際，儘管公司估值很高，但仍可能影響投資者。由於先前與國防部在大規模監控和自主武器問題上存在分歧，公司與政府之間的關係早已緊張。

Internationally, the U.S. government's ability to shut down these models has led the European Union to seek more independence. French President Emmanuel Macron and other officials highlighted that Europe is vulnerable because it controls less than 5% of global AI infrastructure. Consequently, there are plans for Europe to create a coalition of 'AI middle powers' to negotiate better access to advanced models. At the G7 summit, CEO Dario Amodei called for international cooperation to keep regulations consistent, while President Trump suggested that future regulations would depend on how well companies cooperate.

在國際上，美國政府關閉這些模型的能力促使歐盟尋求更多獨立性。法國總統馬克龍和其他官員強調，由於歐洲掌控的全球 AI 基礎設施不足 5%，因此處於脆弱地位。因此，歐洲計劃建立一個「AI 中等強國」聯盟，以協商獲取更先進模型的權限。在 G7 峰會上，執行長 Dario Amodei 呼籲國際合作以保持監管一致，而川普總統則暗示，未來的監管將取決於企業的合作程度。

Currently, foreign access remains blocked. However, the administration has indicated that it no longer views Anthropic as a security threat now that the company has complied with the rules.

目前，外國訪問仍處於封鎖狀態。然而，政府已表示，由於公司已遵守相關規則，目前不再將 Anthropic 視為安全威脅。