在金融市场围绕人工智能乐观派与怀疑派的争论之外，一种关注其灾难性风险的第三种情绪正在升温，包括失控模型引发的网络黑客、生物武器扩散乃至人类灭绝危机。经济学家泰勒·考恩（Tyler Cowen）呼吁悲观主义者通过金融市场交易来为其末日论断“下注押注”。目前，较温和的风险已在资产价格中得到体现：受AI驱动的网络攻击担忧推动，网络安全领军企业群击公司（CrowdStrike）的股价在过去六个月中翻了一倍以上，其对应未来12个月的预期市盈率已飙升至近200倍的高位。

针对更激进的看空策略，投资者可以考虑做空易受AI欺诈与网络渗透的薄弱环节，例如医疗、金融、电信和交通等行业，或购买更脆弱的中小企业及区域性银行的信用违约互换（CDS）。然而，通过传统保险市场转移AI风险的路径正严重受阻：行业标准制定机构保险服务办公室（ISO）今年发布了将生成式AI造成的损失排除在多类理赔之外的新模板，多家大型保险商亦向监管申请免除此类新型威胁责任，而试图填补市场空白的初创保司体量依然微不足道。

保险与对冲机制失效的深层原因在于，AI威胁属于缺乏历史数据和定量概率的根本不确定性，而非可精算定价的传统“风险”。在极端的人类灭绝或AI夺权情景下，此类灾难在金融学上本质上是无法对冲的。投资者不仅无法找到能够定价此类尾部风险的交易模型，更面临灭顶之灾下交易对手方无法存续并履约赔付的荒谬悖论，使得任何末日看跌期权在现实中都沦为空谈。

Beyond the binary debate between AI optimists and sceptics, a third investor sentiment is emerging around existential AI perils, ranging from misaligned hacking models to rogue bioweapons and human extinction. In response, economist Tyler Cowen challenged alarmists to express their convictions through financial markets. Certain milder downstream threats are already being aggressively priced into valuations: mounting fears over AI-assisted cyber warfare have propelled cybersecurity leaders like CrowdStrike, whose share price has more than doubled over the past six months to reach a forward price-to-earnings multiple of nearly 200 times forecast earnings.

For more adventurous bears, short strategies could target sectors highly susceptible to AI-driven fraud and disruption—such as healthcare, finance, telecommunications, and regional banks—potentially through credit-default swaps on vulnerable mid-sized issuers. However, conventional hedging through the insurance industry remains severely constrained. The Insurance Services Office recently introduced contract templates excluding generative AI damages across standard policy categories, and major underwriters are lobbying regulators to exclude novel AI liabilities entirely, leaving only small startups with negligible underwriting capacity attempting to plug the coverage void.

The fundamental impediment to financialising AI catastrophes lies in the distinction between quantifiable economic risk and radical uncertainty, as novel algorithmic perils lack historical actuarial baselines for accurate pricing. At the extreme boundary of existential disaster or superintelligent subjugation, these threats are intrinsically unhedgeable. Investors face the insurmountable paradox that in a true civilisational collapse, no surviving counterparties would exist to settle claims or honour default contracts, rendering traditional tail-risk market hedging completely obsolete.

Source: You can’t trade your way through the AI apocalypse

Subtitle: It’s hard to price the threats posed by a misaligned superintelligence

Dateline: Sep 24th 2026\n