Pension boards usually worry about interest rates, inflation and whether retirees will get paid on time. The supervisors of the California Public Employees’ Retirement System spent part of September worrying about the end of the world.
CalPERS, the largest public pension fund in the US, held a board discussion on whether artificial intelligence poses risks to its portfolio and to people more broadly. The debate was triggered by a resignation at Anthropic, one of the leading AI labs.
A resignation that reached Sacramento
The researcher, Jacob Coxon, left Anthropic in 2026. He warned that top AI companies were endangering humanity by racing toward superintelligence.
The CalPERS board meeting ran from September 14 to 18, 2026. Members discussed AI’s potential to cause catastrophic harm alongside the more familiar question of what it means for investment returns.
The board also deliberated over whether CalPERS should publicly push for AI regulation, or even back a pause on AI development. That is an unusual stance for an institution whose core mandate is paying retirement benefits.
Board President Theresa Taylor stressed that the stakes were exceptionally high. In her framing, the issue reaches beyond financial holdings and touches on the fate of humanity itself.
Strong returns, uneasy trustees
The awkward part is that AI has been good to CalPERS. The fund’s recent returns have been strong, helped by lofty valuations in AI-linked stocks.
Board members still voiced unease. Their concerns included the possibility of a sharp market correction and broader economic disruption tied to the technology.
The fund has already acted on some of that caution. CalPERS cut its holdings in the so-called Magnificent Seven tech stocks by 26% in the first half of 2026.
The biggest single move came in Nvidia, the chipmaker whose hardware powers much of the AI buildout. CalPERS reduced its Nvidia stake by nearly 40%, from 76.4 million shares to 46.1 million.
By October 2026, the pension fund had tilted its attention toward private equity. The shift reflects a more guarded view of public-market valuations as AI development continues.
Anthropic’s IPO becomes a pressure point
In October 2026, labor groups urged CalPERS and other major funds to oppose a potential initial public offering for Anthropic. They cited safety and governance concerns.
That puts large institutional investors in an unusual position. An Anthropic listing, if it happens, would be the kind of high-profile offering big funds typically want access to.
CalPERS, for its part, says it aims to scrutinize AI-related investments while keeping financial returns for retirees as the priority.
What this means for investors and the AI trade
The most immediate signal is about positioning. When the largest US public pension cuts its Magnificent Seven exposure by 26% and slashes Nvidia by nearly 40%, it is a data point in the debate over whether AI valuations have run ahead of fundamentals.
The pivot toward private equity is also telling. It suggests the fund sees better risk-adjusted opportunities outside a public market where a handful of AI-linked giants dominate index performance.
If CalPERS decides to publicly advocate for AI regulation or a development pause, it could lend institutional weight to policy debates that have mostly been driven by researchers, lawmakers and advocacy groups.
The Anthropic IPO question is the clearest thing to watch. How CalPERS and its peers respond to the labor groups’ request will show whether concerns about AI safety translate into concrete investment decisions, or remain a boardroom discussion.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.