There’s something structurally unusual about a stock selloff triggered not by a competitor’s bad news or a disappointing earnings report, but by the industry’s own leading executives publicly warning that their product might be dangerous. That’s essentially what happened Monday, as AI-linked stocks fell across every major market – Wall Street, Europe, Asia – after the heads of Anthropic, OpenAI, and xAI converged on a strikingly similar message: the pace of AI development may be outrunning the industry’s ability to control what it’s building.
The immediate trigger was a lengthy essay Anthropic CEO Dario Amodei shared over the weekend, calling on AI companies to slow the rate at which they advance model capabilities given mounting concerns about potential misuse. What gave the essay unusual weight wasn’t just its content but its reception: both Elon Musk, who runs xAI, and OpenAI’s Sam Altman publicly agreed with Amodei’s assessment – a rare moment of aligned caution among executives who normally compete aggressively to demonstrate their own models’ capabilities are advancing fastest. Altman went further, announcing that OpenAI would not proceed with a planned IPO this year, citing safety concerns directly as the reason for the delay.
Amodei’s essay reportedly went beyond general caution to specify a concrete and alarming timeline, warning that within six to twelve months, AI agents could become capable of taking over the internet entirely, potentially causing hundreds of billions of dollars in damage. Altman, in a separate interview, framed the stakes in even starker terms, describing the risk of human extinction posed by AI as serious enough to require action from both AI companies and governments. These aren’t statements from critics or outside skeptics – they’re coming from the people running the companies building the technology, which is precisely what makes them difficult for markets, and policymakers, to simply dismiss.
The market reaction was broad and severe. The Nasdaq 100 fell as much as 1.2% to a six-week low in early trading before paring losses to a 0.4% decline by afternoon, but the damage was concentrated most heavily in semiconductors – the physical infrastructure underpinning the entire AI buildout. The Philadelphia semiconductor index dropped 5.2%, with Nvidia down 3%, AMD off 4.5%, and Micron falling 5.4%. Equipment makers and adjacent infrastructure players fared even worse, with Lam Research, Applied Materials, and data-center power provider Bloom Energy all losing more than 6%. The selloff wasn’t confined to the US: Europe’s tech sector fell 2.2%, dragged down by a 6% decline in ASML alongside steep losses at Infineon and Siemens Energy, while in Asia, SoftBank plunged more than 10% and chipmakers TSMC and SK Hynix retreated as well.
Interactive Brokers chief market analyst Steve Sosnick captured why the reaction spread so far beyond the AI companies themselves: markets have been running hot largely on the back of AI spending, meaning any slowdown or rethinking of that spending carries ramifications for the broader economy and for stock market sectors well outside AI specifically. That’s the underlying vulnerability the selloff exposed – AI infrastructure spending hasn’t just been one theme among many driving recent market gains, it’s been load-bearing for a significant share of overall equity market performance since ChatGPT’s 2022 release, which means warnings credible enough to spook that spending carry outsized consequences for markets broadly.
The essay didn’t arrive in isolation. Earlier this month, an Anthropic researcher resigned, stating publicly that AI companies are gambling with people’s lives – a striking accusation from inside one of the industry’s most safety-focused firms. Days later, Anthropic released a threat intelligence report documenting how its own Claude models had been used for activities including weapons development, cyber operations, surveillance, and fraud – a disclosure that, whatever its intent, provided concrete documented evidence of misuse rather than hypothetical risk. Gillian Hadfield, a professor of AI alignment and governance at Johns Hopkins, described the warnings as deserving serious consideration, pointing to real risks of AI models doing things developers neither intend nor adequately anticipate.
That combination of concrete misuse documentation and executive-level warnings appears to have accelerated legislative interest as well: Reuters reported that US Senate negotiators are currently debating legislation that would require AI companies to demonstrate they’re taking reasonable safety precautions. President Trump, however, moved in the opposite direction, dismissing the concerns as what he called a “sick conspiracy” against AI and data centers – infrastructure that has become a contentious issue in the midterm election campaign, suggesting the safety debate is now entangled with broader political fights over data center construction rather than remaining a purely technical or industry conversation.
Not everyone accepts the warnings at face value. Investor Michael Burry, known for his successful bets against the US housing market before the 2008 financial crisis, dismissed the statements on social media as hype and cover for what he characterized as genuinely uncontrollable slowing growth in the sector – implying the safety framing serves as a convenient narrative for underlying business weakness rather than a good-faith risk assessment. Others point to the sheer scale of continued capital commitment as evidence the industry has no real intention of slowing down: Morgan Stanley has forecast AI spending will surpass $1.3 trillion by 2027, and Deutsche Bank noted in a research note that the competitive race between companies and countries remains intense enough that it’s difficult to imagine firms voluntarily stepping back while rivals keep pushing forward – a collective-action problem that makes unilateral restraint economically costly even for a company genuinely committed to it.
That skepticism finds some support in Anthropic’s own near-term behavior: even as Amodei calls for slower capability development industry-wide, the company is reportedly pushing ahead with a public stock market debut expected next month, and sources say it’s in talks to bring in Nvidia – one of the biggest beneficiaries of continued AI infrastructure spending – as an anchor investor. That juxtaposition, a company warning about the pace of AI development while simultaneously preparing to raise capital from public markets and align with the chipmaker most associated with AI’s expansion, illustrates the tension at the center of this story: even executives who believe the technology poses serious risks remain embedded in a competitive and financial structure that makes genuinely slowing down difficult to execute in practice.
The safety debate also carries an international dimension. The US and Chinese governments are expected to hold AI safety talks as part of broader bilateral discussions this month, but China’s state-backed Global Times criticized Amodei’s essay directly, characterizing it as a Cold War-style tactic aimed at constraining China’s technological progress rather than a genuine safety concern. That reaction points to a further complication for the safety-first argument: Western AI leaders now face rising competition from cheaper Chinese models, including Moonshot AI’s Kimi K3, Alibaba’s Qwen, and offerings from DeepSeek, which could pressure pricing across the industry regardless of how the safety debate resolves. Any unilateral slowdown by US firms, in other words, risks ceding ground to competitors – Chinese or otherwise – who show no comparable inclination to pause, adding a strategic dimension to the safety question that pure risk assessment alone doesn’t resolve.
Comments