在職場上，最令人心累的莫過於被排擠。明明自己沒做錯事，甚至算得上仆心仆命，但卻被針對、被報復，當中感受只有過來人才明白。AI巨頭Anthropic就於2026年初經歷了這令人心累的困局，最後竟然可以化險為夷，浴火重生。回顧Anthropic的90天翻生歷險記：
🐣 認識Anthropic：OpenAI 唯一忌憚的宿敵
Anthropic 的創辦人 Dario Amodei 兄妹與其他 9 位頂級科學家，原本是 OpenAI 的核心靈魂人物。2021 年底，Dario 團隊因不滿 OpenAI 為了急著商業化而將「AI 安全審查（Alignment）」降級，選擇集體憤而離職。他們創立 Anthropic 後，首創了「憲法 AI（Constitutional AI）」訓練法，也就是給 AI 一套基本原則，讓 AI 透過「自我反思」來確保自己不會失控、不會有害。後來，包括 Google、Meta 甚至 OpenAI 自己的後期模型，都全面借鑑了這套安全技術。
💥 拒絕做聽話的孩子：與美國國防部起衝突
原本與美國國防部簽有一筆價值 2 億美元的技術合約。但從 2026 年初開始，國防部與軍方要求Anthropic允許軍方對 Claude 進行「任何合法用途（any lawful use）」的無限制使用。Anthropic 拒絕，並明確畫出兩大不可侵犯的安全紅線：
（一）拒絕用於「全面性國內監視（Mass Domestic Surveillance）」
（二）拒絕用於「完全自主的致命武器系統（Fully Autonomous Lethal Weapons）」（即由 AI 自行決定開火殺死人類，而非由人類軍人控制）。
🚫 美國政府下達封殺令
特朗普於社群平台上痛罵 Anthropic 是由「左翼瘋子」組成的公司，無視國家安全。隨後五角大廈將 Anthropic 定性為「國家安全供應鏈風險（Supply Chain Risk）」，這項頭銜通常只會用在華為或俄羅斯卡巴斯基等「敵國企業」身上，特朗普政府下令所有聯邦政府機構（包括財政部等非軍事單位）以及所有的國防承包商，必須在六個月內全面停用、淘汰 Anthropic 的產品。
🌈 逆襲成功，重修舊好
4 月份 Anthropic 帶著尚未公開的 Claude Mythos 秘密簡報特朗普政府時，局勢瞬間逆轉。特朗普接著在接受 CNBC 訪問時，態度與兩個月前截然不同，風向立刻大轉變，直接在電視上用「They are very smart」和「I like smart people」向 Anthropic 示好。了解更多Claude Mythos的逆天實力👉https://softmap.com.hk/news-software-20260420/
📣 Anthropic近況：將影響力擴展至商界
5月份宣布推出10個新的AI智能體 (AI Agents)，旨在幫助華爾街銀行和保險公司實現工作自動化，這些工具開箱即用，當中包括製作推銷簡報（pitchbooks）、結帳（closing the books），以及起草信用備忘錄。高盛、花旗集團、Visa、美國國際集團（AIG）等公司也已採用Claude平台。
在職場上，你是不是也聽過這些質疑？「你不合群、你破壞規矩、你成不了大事……」如果你正在經歷被排擠或不被看好的低谷，請看看這家 AI 巨頭 —— Anthropic（Claude 的母公司）的經歷。它用現實告訴全世界：在絕對的「實力」面前，一切的政治、針對和唱衰，都只是砂石。實力，就是你最強的武器！希望這個故事可以帶給你勇氣面對當前的困境。
In the workplace, nothing is more draining than being sidelined. When you have done nothing wrong and have even poured your heart and soul into your work, yet still find yourself targeted and retaliated against—the pain is something only those who have lived through it truly understand. AI giant Anthropic experienced this exact exhausting dilemma in early 2026, only to miraculously turn the crisis into an opportunity and rise from the ashes. Let’s look back at Anthropic’s 90-day resurgence journey:
🐣 Meet Anthropic: OpenAI’s Only Feared Rival
The founders of Anthropic, siblings Dario and Daniela Amodei, along with 9 other top scientists, were originally core, foundational figures at OpenAI. In late 2021, Dario’s team chose to resign collectively out of frustration with OpenAI deprioritizing “AI safety alignment” in its rush to commercialize. After founding Anthropic, they pioneered the “Constitutional AI” training method—a framework that gives AI a core set of principles, enabling it to ensure it won’t spiral out of control or cause harm through “self-reflection.” Eventually, tech giants including Google, Meta, and even OpenAI themselves thoroughly drew inspiration from this safety framework for their later models.
💥 Refusing to Be a Compliant Child: Clashing with the US Department of Defense
They originally held a technology contract with the US Department of Defense worth $200 million. However, starting in early 2026, the Department of Defense and the military demanded that Anthropic allow unrestricted use of Claude for “any lawful use.” Anthropic refused and firmly drew two non-negotiable safety red lines:
(1) Absolute refusal to be used for “Mass Domestic Surveillance.”
(2) Absolute refusal to be used for “Fully Autonomous Lethal Weapons Systems” (where the AI independently decides to fire and kill humans, rather than being controlled by a human soldier).
🚫 The US Government Standardizes a Full Ban
Trump lashed out at Anthropic on social media, branding them a company run by “left-wing nut jobs” who ignore national security. Following this, the Pentagon designated Anthropic as a “National Security Supply Chain Risk”—a label typically reserved only for “foreign adversarial entities” like Huawei or Kaspersky. The Trump administration subsequently ordered all federal agencies (including non-military sectors like the Treasury Department) and all defense contractors to completely phase out and eliminate Anthropic products within six months.
🌈 A Successful Counterattack and Rebuilding Ties
In April, the tables turned instantaneously when Anthropic brought its unreleased, next-generation model, Claude Mythos, to a secret briefing with the Trump administration. Shortly after, during an interview with CNBC, Trump’s stance shifted entirely from his tone two months prior. The political wind reversed immediately as he publicly gestured goodwill toward Anthropic on television, stating “They are very smart” and “I like smart people.”
👉 Learn more about the nuclear-level capabilities of Claude Mythos here:https://softmap.com.hk/news-software-20260420/
📣 Anthropic’s Recent Movements: Expanding Influence Across the Business Sector
In May, the company announced the rollout of 10 new out-of-the-box AI Agents designed to automate intensive workflows for Wall Street banks and insurance firms. These tasks include constructing pitchbooks, closing the financial books, and drafting credit memos. Global financial powerhouses like Goldman Sachs, Citigroup, Visa, and AIG have already onboarded and adopted the Claude platform.
Have you ever heard similar doubts or criticisms levelled against you in your own career? “You’re not a team player… You’re breaking the rules… You will never achieve anything great…” If you are currently navigating a professional low point, being targeted, or facing underappreciation, look closely at what AI titan Anthropic just overcame. Their journey proves to the world that in the presence of absolute “strength and competence,” all political games, targeted attacks, and cynical voices are merely dust in the wind. Excellence is your ultimate weapon! We hope this story brings you the courage needed to face your current workplace challenges head-on.
參考資料 Reference:
MicroVentures https://microventures.com/microventures-portfolio-company-anthropics-history-and-milestones
Anthropic https://www.anthropic.com/research/constitutional-ai-harmlessness-from-ai-feedback
BBC https://www.bbc.com/news/articles/cvg3vlzzkqeo
The San Francisco Standard https://sfstandard.com/2026/02/27/anthropic-trump-blacklist/
CNBC https://www.cnbc.com/2026/02/27/trump-anthropic-ai-pentagon.html
BBC https://www.bbc.com/news/articles/cn48jj3y8ezo
CNBC https://www.cnbc.com/2026/04/21/trump-anthropic-department-defense-deal.html
The Wall Street Journal https://www.wsj.com/tech/ai/anthropic-releases-new-ai-agents-for-financial-services-firms-e2829b37