Steve Eisman, the investor whose actions inspired a character in the film The Big Short, says the recent surge in talk about an imminent AI-driven apocalypse is being encouraged by the very companies that develop artificial intelligence. In an appearance on CNBC, Eisman argued that horror-scenario narratives serve the strategic interest of large AI developers because they can lead to regulation shaped in the firms’ favor.
How the public debate shifted
In recent weeks the public conversation about AI has shifted beyond job losses or resource competition to more apocalyptic themes: some commentators now ask whether AI systems could cause civilizational collapse or even human extinction within a few years. Coverage intensified after Jacob Coxon, who worked at both OpenAI and Anthropic, said he resigned in part because some developers at those companies believe AI could pose existential risks within a short timeframe.
Other developers at the same firms reportedly responded by saying they personally assign roughly a 10 percent chance that systems they work on could cause catastrophic harm within a decade. Leaders of major AI companies have publicly advocated for coordinated slowdowns and common safety measures in development.
Eisman’s view: crisis narratives as a market strategy
Eisman contends that these alarmist narratives are not purely technical or ethical concerns but have a strategic business motive. He says companies such as Anthropic and OpenAI see advantage in provoking calls for regulation that they can help shape. By influencing rulemaking, large firms could secure protective market positions — for example, creating barriers that favor a small number of dominant providers.
Economic dynamics: tokenmaxxing and rising scrutiny of costs
The reporting points to an economic mechanism behind some recent behavior. Many companies encouraged employees to use AI tools broadly, which led to large volumes of prompts and heavy token consumption — a practice sometimes called “tokenmaxxing.” While token usage grew dramatically, that did not always translate into proportional productivity gains. Because AI providers charge based on usage, tokenmaxxing meant substantial bills for client companies.
As a result, many customers have begun to curb employee usage, introduce limits, or move to alternatives like open-weight models that are easier to audit and control. Eisman argues this is worrying for major AI vendors: if customers tighten spending, revenues could decline, and business models that rely on continuous cash flow — models without a safety moat or backup — could become fragile.
Eisman’s investment moves and conclusions
In the CNBC interview Eisman said he has reduced his AI exposure. He added that if firms such as Anthropic and OpenAI truly believed their products were existentially dangerous, they could take more concrete steps than stoking public fear — for example, pausing fundraising rounds or delaying planned public listings. Eisman told the program he has not seen such steps, and therefore he does not believe the AI threat justifies existential alarm for humanity in the near term.
Why this matters
Eisman’s remarks highlight a dynamic worth noting: public debate about technological risk can be shaped by the commercial incentives of the firms involved. The shift from technical safety concerns to doomsday narratives can influence the form of regulation that emerges, and that regulation in turn can reshape market structure and the fortunes of the industry’s players.
Eisman’s perspective does not in itself disprove AI-related risks, but it underlines that strategic business motives may be an important factor behind alarmist messaging — a factor that policymakers and the public should take into account when assessing claims of imminent catastrophe.
