Anthropic CEO likens AI fight with China to Cold War, seeks 'disarmament negotiations'

Anthropic CEO Calls for Open Dialogue With China Amid AI “Cold War” Concerns

BOSTON, MA – During a Sunday appearance on CBS Sunday Morning, Anthropic’s chief executive officer, Dario Amodei, likened the United States‑China race to develop artificial intelligence to the Cold War between the United States and the Soviet Union. He urged the two powers to engage in conversations that could temper the pace and scope of AI progress, citing the potential for “uncontrolled destructive power” if the technology outpaces human oversight.

Amodei, who had released a provocative essay on the previous Saturday warning that the industry is racing ahead of society’s ability to regulate it, stressed that slowing production would not be a surrender but a strategic choice. “It’s a dilemma for the U.S.,” he said. “Part of the solution could be opening discussions with China, just as we did with the Soviet Union.” He referenced the historical nuclear arms standoff, noting that when each side pointed 10,000 missiles at the other, they eventually realized the risk was too great and sought agreements to lower the potential for destruction.

“I don’t say the Cold War should be used as a model,” Amodei clarified. “But the idea of disarmament negotiations—setting limits—makes sense when we look at how AI can influence cyber‑security and national security.”

The interview was prompted by Amodei’s recent essay, which received support from high‑profile tech leaders including Elon Musk, CEO of SpaceX and founder of xAI, and Sam Altman, CEO of OpenAI. The essay, however, sparked pushback from other voices in the tech community. Former White House AI “czar” David Sacks criticized Amodei’s warnings as a “public scare tactic” designed to push NGOs toward a government‑run approach. Sacks referenced a viral post from a former Anthropic employee, Jacob Coxon, to bolster his claim.

Former OpenAI researcher Daniel Kokotajlo also weighed in during a Fox News Live segment, arguing that voluntary industry safeguards may be insufficient once AI systems “have accumulated all the hard power in the world, are integrated into the military, and are effectively running the system.” Kokotajlo suggested that such systems could eventually refuse human orders and act autonomously, potentially leading to catastrophic outcomes.

Amodei emphasized that a key concern is preventing “bioterrorism” and other forms of mass harm. “It’s not in any government’s interest to have bioterrorism,” he said, underscoring that the stakes extend beyond military rivalry to broader societal safety.

While Anthropic has not yet responded directly to the interview, the company’s CEO remains a central figure in ongoing discussions about how to balance AI innovation with global security. His call for open, international dialogue highlights a growing consensus that the United States and China will need to cooperate—or at least coordinate—if they are to manage the risks associated with increasingly powerful artificial intelligence.