美國防部與 OpenAI、微軟、Nvidia 等七巨頭簽署「AI 合作協議」，Anthropic 拒讓步遭川普封殺出局

📄完整原文· 由 trafilatura 自動擷取Gemini 翻譯1532 字

The era of AI warfare has arrived! The Pentagon made a shocking announcement on Friday, confirming it has successfully allied with 7 tech giants including OpenAI, Microsoft, Google, Nvidia, and SpaceX to fully integrate AI into the U.S. military's top-secret networks to accelerate battlefield decision-making. However, Anthropic, which emphasizes "AI ethics and safety," was blocked by the Trump administration and sued for refusing to allow its models to be used for fully autonomous weapons; meanwhile, OpenAI took its place. As AI takes over more military command functions, experts are sounding the alarm: never let machines have the final button to decide life and death! (Previous coverage: U.S. CFTC uses AI to review crypto registrations! Chairman Mike Selig: Filling layoff gaps, efficiency becomes super high) (Background: Florida investigates OpenAI: Gunman asked ChatGPT how to disable safety before opening fire, involved in multiple violent and fatal incidents) The U.S. military is weaponizing Artificial Intelligence (AI) at an unprecedented speed. The U.S. Department of Defense (Pentagon) announced on Friday (the 1st) that it has reached an agreement with seven top tech companies, allowing the military to use their AI infrastructure and models in classified computer networks, with the goal of gaining an overwhelming advantage on the rapidly changing modern battlefield. This impressive list of the military-industrial complex includes: Google, Microsoft, Amazon AWS, Nvidia, OpenAI, Reflection, and Elon Musk's SpaceX. The Department of Defense stated that the resources of these companies will help "enhance the decision-making capabilities of warfighters in complex operational environments." The Pentagon pointed out that military personnel are already using these AI capabilities through an official platform called GenAI.mil, significantly reducing tedious tasks that previously took months to complete to just a few days. Pentagon CTO Emil Michael specifically named Nvidia and the startup Reflection. Both companies provide "open-source AI models," which Michael views as a priority strategy aimed at building a "U.S. alternative" that can compete with China's rapidly developing open-source AI systems. However, in this frenzy between the military and Silicon Valley, the absence of AI giant Anthropic is particularly glaring. According to reports, Anthropic held firm on its ethical bottom line during negotiations. The company demanded explicit guarantees in the contract to ensure that the U.S. military "would not use its technology for fully autonomous weapons and the surveillance of U.S. citizens." This demand thoroughly angered high-level officials in Washington. Secretary of Defense Pete Hegseth took a hardline stance, demanding that the company allow the Pentagon to use the technology for "any lawful purpose." Subsequently, President Donald Trump attempted to ban federal agencies from using Anthropic's chatbot, Claude; Hegseth further attempted to label the company a "supply chain risk." Ultimately, Anthropic chose to take the Trump administration to court. After Anthropic was out, OpenAI quickly reached an agreement with the Pentagon in March of this year, replacing Anthropic with ChatGPT. In a statement on Friday, OpenAI did not shy away from supporting the military: "As we announced when we reached our agreement months ago, we believe that those who defend the United States should have access to the world's best tools." However, insiders also revealed that OpenAI secured a clause in the contract stating that "there must be human oversight when AI acts autonomously." As AI rapidly infiltrates the command structure, ethical and risk issues have surfaced. Helen Toner, Acting Director of Strategy at Georgetown's Center for Security and Emerging Technology (and a former OpenAI board member), warned that while AI is adept at summarizing intelligence and monitoring video, the military must never over-rely on this technology: "There is a phenomenon called 'automation bias,' where people easily assume that machines perform better than they actually do." Drawing lessons from recent cases where Israel relied heavily on AI to mark targets in the conflicts in Gaza and Lebanon, leading to heavy civilian casualties, how to strike a balance between pursuing strategic advantages and upholding human ethical bottom lines will be the most severe test left for all of humanity in the AI era.

資料狀態✓ 已擷取全文閱讀原文（動區 BlockTempo）

🔍歷史類似事件· 關鍵字 + 標的比對6 則

2026-04-30

Bloomberg：Anthropic 正洽談 9000 億美元估值新融資，有望超越 OpenAI

相似度 140%關鍵字 anthropic/openai同分類 zh

2026-04-30

彭博：Anthropic 正洽談 400 億美元估值新融資，有望超越 OpenAI 成最貴 AI 公司

相似度 140%關鍵字 anthropic/openai同分類 zh

2026-04-24

黃仁勳發內部信：強制 1 萬名 NVIDIA 員工改用 OpenAI Codex，GPT-5.5 用 GB200 晶片跑

相似度 140%關鍵字 openai/nvidia同分類 zh

2026-04-24

黃仁勳發全員信擁抱 OpenAI Codex：1 萬多名 NVIDIA 員工已上手，GPT-5.5 跑在 GB200 上

相似度 140%關鍵字 openai/nvidia同分類 zh

2026-04-23

OpenAI 執行長 Sam Altman 砲轟 Anthropic：恐懼行銷 Claude Mythos 只為壟斷 AI

相似度 140%關鍵字 anthropic/openai同分類 zh

2026-04-21

川普鬆口國防部極可能與 Anthropic「達成 AI 合作協議」，大讚：他們表現漸入佳境

相似度 140%關鍵字 anthropic/合作協議同分類 zh

💡 目前用關鍵字 + 標的比對（MVP）· 之後會升級為 embedding 語意搜尋

原始資訊

ID：2d1b266845

來源：動區 BlockTempo

發佈：2026-05-02 02:57:17

分類：zh_news · 導出分類 zh

標的：未指定

社群投票：+0 / −0 · ⭐ 0 重要 · 💬 0 留言