本周软体与金融服务股票出现约 $300 billion 的抛售，被外界认为可能由 Anthropic PBC 发布的新法律产品所引爆；即使不宜把市场崩跌归因于单一触发点，市场对 Anthropic 造成破坏式冲击的焦虑，凸显其看似不可阻挡的生产力。Anthropic 约有 2,000 名员工，却表示仅在 1 月就推出 30 多项产品与功能，并在周四发布 Claude Opus 4.6，主打处理知识工作，直接升高对 Salesforce 与 ServiceNow 等传统 SaaS 公司的压力。相较之下，OpenAI 的人力约为 Anthropic 的 2 倍，而 Microsoft Corp. 与 Alphabet Inc. 的 Google 分别约有 228,000 与 183,000 名员工；但评论指出，Anthropic 在生成程式码与操作电脑的工具上，推出的成果被认为超过这些更大公司近期能交付的产品影响力。

这种效率部分来自一种悖论性的来源：近乎意识形态的使命与安全文化。Anthropic 由前 OpenAI 成员创立，他们认为 OpenAI 对 AI 的安全过于漫不经心，尤其是对人类存亡风险；在 Anthropic，这个议题演变成公司信条，由 CEO Dario Amodei 作为核心象征。Amodei 每月两次召集员工进行所谓的 Dario Vision Quest（DVQ），长谈价值对齐、地缘政治，以及 Anthropic 技术对劳动市场的影响；他在 2025 年 5 月警告，AI 进展可能在未来 1 到 5 年消除多达 50% 的入门级办公室工作，而公司仍以近乎宗教式的热忱推进「安全 AI」，就业安全并未成为其理想的一部分。外部描述也包含「类邪教」氛围：工程主管 Boris Cherny 表示，多数人加入是为了让 AI 变得安全；同时，Meta Platforms Inc. 曾以承诺远离开源 AI 作为挖角话术，因为 Anthropic 员工认为开源可自由修改的模型更具风险。

Amodei 近期又发布约 20,000 字的文章，主张 AI 带来迫近的文明风险，而 Anthropic 也为 Claude 发布冗长的「constitution」，以更清楚的规范处理被关闭的可能性；这既是安全机制，也带有对系统道德地位的哲学主张。其安全执念被认为带来商业信任：根据 Scale.ai 研究者的排名，Anthropic 的模型更诚实，较不易幻觉，且更可能直接承认不知道，因而更吸引企业客户。评论并将其优势归因于使命一致降低内耗，以及聚焦打造 Claude Code 等高效编码工具；Sebastian Mallaby 亦以「为崇高目标而战」的类比，对照 Sam Altman 所领导的 OpenAI 分散押注与「领跑者傲慢」。在此同时，Anthropic 据报正以 $350 billion 估值募资 $10 billion，使成长压力上升并考验其安全文化能否在更高利害下维持，且其产品扩张可能持续撼动市场并影响大量工作。

This week’s roughly $300 billion selloff in software and financial services stocks was widely framed as being sparked by Anthropic PBC and a new legal product it released; even if it is misguided to credit a rout to a single trigger, fear of Anthropic-driven disruption spotlights its seemingly unstoppable productivity. Anthropic has about 2,000 employees yet says it launched more than 30 products and features in January alone, then kept pace on Thursday with Claude Opus 4.6, a model aimed at knowledge-work tasks that turns up pressure on legacy SaaS firms such as Salesforce and ServiceNow. By comparison, OpenAI’s workforce is about twice Anthropic’s, while Microsoft Corp. and Alphabet Inc.’s Google have about 228,000 and 183,000 staff; the commentary argues Anthropic’s tools for code generation and computer operation have exceeded the recent product impact of those larger companies.

That output is tied partly to a paradoxical source: a mission-and-safety culture that functions like an ideology. Anthropic was founded by ex-OpenAI staff who thought OpenAI was too casual about safety, especially existential risk; inside Anthropic the concern becomes a creed, with CEO Dario Amodei as its central figure. Twice a month Amodei runs a “Dario Vision Quest” (DVQ) to talk about value alignment, geopolitics, and labor-market effects; in May 2025 he warned AI could erase up to 50% of entry-level office jobs within one to five years, while the company’s near-religious zeal for “safe AI” appears willing to accelerate that outcome and does not treat employment safety as part of the ideal. Accounts also describe a cult-like atmosphere: lead engineer Boris Cherny says people are there to make AI safe, and Meta Platforms Inc. tried to recruit by promising to move away from open-source AI because Anthropic staff see freely modifiable systems as dangerous.

Amodei has also published an essay of about 20,000 words on imminent civilizational risk, and Anthropic released a lengthy “constitution” for Claude that sets clearer rules for handling the possibility of shutdown; it is presented as both a safety mechanism and a claim about moral status. The safety obsession is depicted as commercially valuable: a Scale.ai researchers’ ranking says Anthropic models are more honest, less prone to hallucination, and more likely to admit uncertainty, helping win more enterprise clients. The commentary attributes advantage to mission alignment reducing internal friction and to focus on tools such as Claude Code; Sebastian Mallaby compares it to fighting for a noble cause and contrasts it with Sam Altman’s OpenAI pursuing many avenues and suffering “front-runner” arrogance. Meanwhile, Anthropic is reportedly raising $10 billion at a $350 billion valuation, increasing growth pressure and testing whether its safety culture holds as the stakes rise, even as its expansion continues to rattle markets and potentially many jobs.