Anthropic is planning an IPO this fall.[1] In an unconventional approach to pre-IPO corporate communications, Evan Hubinger of Anthropic tweeted on September 8 that:[2]
… we really do earnestly believe AI could kill all humans!
Given these high stakes, how should we think about Anthropic’s IPO? The implicit argument is that AI is so dangerous that Anthropic must raise vast sums of money to safely develop an AI that does not lead to human extinction. The AI singularity is coming whether we like it or not, and thus we need guardrails prudently installed by Anthropic. This argument is ESG on steroids: Extinction, Singularity, Guardrails.
Jones (2026), in a balanced discussion, provides the following comparison for the upsides and downsides of AI:
In anticipation of the first test of the atomic bomb, the scientists of the Manhattan Project faced potentially catastrophic risks: for example, what if the nuclear chain reaction continued unabated, igniting the atmosphere and potentially killing all life on Earth? … They estimated the probability to be extremely low, and the Trinity test went forward. But how large would the risk have to be to avert the test?
… Suppose on one side that AI has incredible benefits, raising economic growth to 10 percent per year and thus doubling the average standard of living every seven years. However, it comes with a one-time risk of killing everyone on the planet.
Anthropic’s IPO is like Robert Oppenheimer doing an IPO for the Manhattan Project in 1945.
Whether Anthropic is proceeding wisely is a complicated issue with valid arguments on both sides. Instead of tackling this thorny dilemma, let me ask the following narrow question: if you have a company whose operations may cause death and destruction, does the danger increase if that company has an IPO?
The good news is that there are reasons to believe that public firms act safely compared to private firms because public firms face more scrutiny and have more financial resources to spend on safety. The bad news is that as an empirical matter, public firms sometimes have more workplace accidents than private firms.
Let’s consider one industry that poses many safety concerns: coal mining. Gilje and Wittry (2021) study U.S. coal mines from 1985 to 2018, and find that:
… workplace safety is strongly related to listing status. Specifically, mines that are purchased by public firms or are part of an IPO experience a large increase in safety violations. However, this increase is concentrated in what appears to be tail risk, or violations for potential accidents stemming from low-probability events. Such violations increase by nearly 57%. Furthermore, upon a listing status change, workplace fatalities increase four-fold relative to the unconditional sample mean.
We can also consider specific IPOs of coal mining firms. International Coal Group (ICG) had an IPO on December 12, 2005. Three weeks later, its Sago Mine in West Virginia exploded, killing 12 miners. Like Anthropic, ICG was touted as being especially safety conscious. Subsequent lawsuits alleged that ICG “wrongly claimed in its initial public offering for stock purchasers that its operations had a good safety record.”[3]
A more striking case involves the Ulyanovskaya mine disaster in 2007. In preparation for an IPO, the Russian firm Yuzhkuzbassugol brought an outside auditor to inspect its supposedly safe and modern mine in Siberia.[4] The auditor, along with senior mining executives, went underground to observe a demonstration of the mine’s “advanced mining safety system.” A massive methane explosion occurred, killing more than 100 people including the auditor and the management team.
What lessons can we learn from the Ulyanovskaya disaster? First, just because management thinks that something is safe, that doesn’t make it safe. Second, even when management is aware of the risk and devotes considerable resources to reducing it, catastrophe can still occur.
I’ve previously discussed many aspects of IPOs: whether you should buy them, why a company might do one, and why they occur during stock market bubbles. The Pets.com IPO of February 2000 is now widely seen as marking the beginning of the collapse of the tech stock bubble. Let’s hope that future historians do not look back at the Anthropic IPO and see it as marking the beginning of the collapse of human civilization.
Endnotes
[1] References to this and other companies should not be interpreted as recommendations to buy or sell specific securities. Acadian and/or the author of this post may hold positions in one or more securities associated with these companies.
[3] “Union says ICG Misled Investors,” The Wall Street Journal, June 18, 2008.
References
Gilje, Erik P., and Michael D. Wittry. Is public equity deadly? Evidence from workplace safety and productivity tradeoffs in the coal industry. No. w28798. National Bureau of Economic Research, 2021.
Jones, Charles I. "AI and Our Economic Future." Journal of Economic Perspectives, vol. 40(3), pages 3-22, Summer 2026.
