虑Anthropic PBC计划中的大规模首次公开募股（IPO）的投资者们，正面临一个不同寻常的难题：如何为一家可能被某些人认为会帮助终结人类文明的公司进行估值。
Investors considering Anthropic PBC’s planned mega-IPO are grappling with an unusual conundrum: How to value a company that some fear could help wipe out the human race.
在X平台上，有研究人员发文称人工智能终结人类文明的可能性达到10%或更高，这一言论引发了一场讨论，Anthropic的联合创始人兼首席执行官达里奥·阿莫迪也对此发表了看法。他在文章中阐述了相关担忧：人类可能会失去对人工智能系统的控制，这些工具可能被滥用用于网络攻击和生物恐怖主义，而且如果顶尖研究人员不放缓开发进程，还可能造成严重的经济动荡。
Researchers posting on X about a 10% chance or greater that AI would wipe out humanity sparked a discussion that Dario Amodei, Anthropic’s co-founder and chief executive officer, weighed in on. His essay outlined concerns that humans could lose control of AI systems, the tools could be misused for cyberattacks and bioterrorism, and could create serious economic disruption if top researchers don’t slow development.
尽管观察人士对“一家公司可能因导致人类灭绝而承担法律责任”这一观点嗤之以鼻，甚至认为无法在首次公开募股的信息披露中对此进行量化，但Anthropic及其竞争对手OpenAI所面临的风险，与处理危险材料或军事装备的公司所面临的风险类似。
While observers scoffed at the idea that a company could be held liable for killing off the species — or even quantify it in an initial public offering disclosure — the risks to Anthropic and its rival OpenAI resemble those faced by companies handling hazardous materials or military hardware.
据Slow Ventures合伙人萨姆·莱辛称，2008年金融危机后的处理方式或许能为应对人工智能风险提供借鉴。这位风险投资家提出，美国政府应强制人工智能实验室存入1万亿美元的现金至第三方托管账户，用于未来可能需要的清理工作，这与危机后监管机构要求银行持有更多资本以覆盖损失的做法类似。他甚至表示，这一金额可能仍不足够。
A precedent that could inform handling AI risks is the aftermath of the 2008 financial crisis, according to Slow Ventures partner Sam Lessin. The venture capitalist has proposed the US government should force AI labs to put up $1 trillion in cash in escrow to pay for the future potential cleanup, similar to how regulators demanded after the crisis that banks should hold more capital to cover losses. Even that may not be enough, he said.
莱辛在接受彭博新闻社电子邮件采访时说：“这些责任可能给社会带来2万亿美元甚至4万亿美元的风险，这甚至还没有考虑到他们所说的大规模灭绝问题，这意味着这些公司的总体价值可能低于零。”
“The liabilities might be a $2 trillion or $4 trillion risk to society, even before you consider the mass-extinction they talk about, making the companies worth less than zero overall,” Lessin said in an email interview with Bloomberg News.
不过，莱辛指出，一些投资者或许仍愿意购买Anthropic的股票，并忽略这一尚未明确但可能极其巨大的成本，因为他们相信最终会有其他人来承担这笔费用。Anthropic的发言人拒绝发表评论。
Still, some investors may be willing to buy Anthropic stock and overlook an undefined but potentially enormous cost based on the theory that someone else will pay the bill, Lessin said. A spokesperson for Anthropic declined to comment.
Anthropic计划在其首次公开募股（IPO）中筹集与SpaceX创纪录的首次上市相当甚至更多的资金，尽管围绕人工智能安全的争论已变得近乎末日论，但这一计划仍引发了质疑和嘲讽。然而，这一问题似乎并未对SpaceX的股价造成负面影响。
Anthropic’s plan to raise as much or more in its IPO as SpaceX did in its own record-breaking debut, even as the debate around AI safety turned apocalyptic, has drawn skepticism and mockery. Yet the issue doesn’t appear to have hurt SpaceX shares.
万亿富翁埃隆·马斯克的火箭与卫星公司曾将大量雄心寄托于人工智能，其Grok AI模型也与Anthropic和OpenAI各自推出的产品展开竞争。该公司在IPO文件中并未提及人工智能可能带来的生存性威胁，而是重点指出了虚假信息及“人工智能欺骗”等风险。目前，SpaceX的股价已高于其6月份862亿美元IPO的发行价超过20%。
Trillionaire Elon Musk’s rocket and satellite company pinned much of its aspirations on AI, and its Grok AI model competes with Anthropic and OpenAI’s respective offerings. The company’s IPO filings made no mention of existential threats from AI, instead flagging risks such as misinformation and “AI deception.”SpaceX shares are trading more than 20% above the offering price of its $86.2 billion IPO in June.
InvestorPlace首席技术分析师卢克·兰戈表示，特朗普总统近期与包括阿莫迪在内的科技巨头领导人达成的“具有道德约束力”的人工智能安全协议，缓解了投资者对这一广泛问题的部分担忧。
President Trump’s recent “morally binding” accord on AI safety with big tech leaders, including Amodei, calmed some investor anxiety around the broader issue, said Luke Lango, chief technology analyst at InvestorPlace.
投资处于后期成长型私营公司的Private Shares基金管理董事兼投资组合经理凯文·莫斯表示：“本届政府已经公开表达了担忧，但与此同时，政府也强调放松管制，并且正与中国展开激烈竞争。”
“This administration has come out and expressed concern, but this is also an administration that puts emphasis on deregulation and is in this race with China,” said Kevin Moss, the managing director and portfolio manager of the Private Shares Fund, which invests in late-stage private growth companies.
莫斯指出，人工智能显然需要一个监管框架，但围绕安全性的讨论不太可能打乱Anthropic的IPO计划。
AI clearly required a regulatory framework, but the safety discussion was unlikely to upset Anthropic’s IPO plans, Moss said.
迄今为止，减轻安全风险的压力主要落在人工智能公司自身身上，而投资者甚至可能欢迎企业在解决这一问题上进行一定程度的投入。
So far, the pressure to mitigate safety risks rests primarily with AI companies themselves, and investors may even welcome a certain amount of spending to address it.
Pitchbook负责后期公司的高级研究分析师哈里森·罗尔费斯表示，新的安全机制虽然会增加Anthropic的成本，但可能提升客户对其产品的信任，并在中国或开源竞争对手无法达到相同标准时，帮助公司构建竞争护城河。
New safety mechanisms will increase Anthropic’s costs but could enhance customer trust in its products and help create a competitive moat if Chinese or open-source competition is unable to match the same standards, said Harrison Rolfes, Pitchbook’s senior research analyst for late-stage companies.
尽管该公司预计拥有保险，但历史上保险覆盖范围和争议问题常常导致股东在灾难发生时面临巨额账单。2010年，BP的“深水地平线”钻井平台爆炸并沉没，造成11人死亡，并引发了美国历史上最严重的海上石油泄漏事故。当时，该公司所谓的“专属保险”安排上限仅为7亿美元，而最终成本却高达数百亿美元。
Though the company is expected to have insurance, coverage limits and disputes have historically left shareholders with a large bill in disasters. When BP’s Deepwater Horizon oil rig exploded and sank in 2010, killing 11 people and causing the biggest offshore oil spill in US history, the company’s so-called captive insurance arrangement was limited to $700 million. The cost climbed into the tens of billions of dollars.
据专注于人工智能和航天等新兴行业的保险公司Relm Insurance的创始人兼首席执行官乔·齐奥科夫斯基（Joe Ziolkowski）称，保险行业一直通过传统合同为人工智能公司提供保障，并在存在保障空白的领域开发新产品。他表示，由于缺乏历史数据以及对人工智能责任形式的不确定性，为人工智能企业找到合适的保险仍然是一项挑战。
The insurance industry has been providing coverage to AI companies through traditional contracts and creating new products where gaps existed, according to Joe Ziolkowski, founder and CEO of Relm Insurance, an insurer focused on emerging industries such as AI and space. Finding coverage remained a battle for AI businesses given the lack historical data and uncertainty as to how AI liability would present itself, he said.
齐奥科夫斯基指出，随着人工智能从单纯提供信息转向做出决策并与其他智能体互动，当损失不可避免地发生时，谁应承担法律责任这一问题正变得愈发紧迫。
As AI moves from just providing information to making decisions and interacting with other agents, the issue of who’s legally responsible when losses inevitably arise is becoming more pressing, Ziolkowski said.
然而，由于在人工智能智能体失控方面的法律判例几乎不存在，因此考虑以高达2万亿美元估值投资Anthropic的投资者实际上面临一个二选一的困境：要么接受该公司可能为某些最极端的人工智能场景承担全部责任的风险，要么干脆不投。
With little direct precedent among legal judgments for AI agents going rogue, however, investors considering buying into Anthropic at a valuation of as much as $2 trillion face what amounts to a binary choice – accept the risk that the company may be fully on the hook for some of the most stark AI scenarios, or don’t.
齐奥科夫斯基表示：“通过公开文件，你可以大致了解这些公司在合规、保险或自我保险方面投入了多少资金，但投资者不太可能因为这些无法量化的风险而降低对这些IPO的估值。”
“You can have an idea when you are looking at public filings how much these companies are putting into compliance or insurance or self-insuring, but it is unlikely that investors are going to discount these IPOs for exposures they can’t quantify,” Ziolkowski said.
“在一些问题尚未能够被客观量化之前，现在是筹集资本的好时机。”
“It is a good time to raise capital before some of these things become objectively quantifiable.”