Anthropic CEO Warns U.S.–China AI Race Resembles Cold War, Calls for “Disarmament Negotiations”

This new interview builds directly on the concerns highlighted in our earlier report on Udumbara.net, which documented how Anthropic’s Claude models were exploited for weapons research, cyber operations, and surveillance — including activity linked to the Chinese Communist Party (CCP). Read Part 1 here: https://udumbara.net/anthropic-reveals-global-misuse-of-claude-ai-china-at-the-center-of-military-and-intelligence-operations

A Cold War Analogy: “Potential Destruction” on Both Sides

Speaking on CBS Sunday Morning, Amodei said the U.S.–China AI competition has reached a point where both nations must confront the risk of “potential destruction” from each other’s rapidly advancing systems. He compared the situation to the U.S.–Soviet nuclear standoff, noting that even bitter rivals eventually recognized the need to limit escalation.

“We each had 10,000 nuclear missiles pointed at each other,” Amodei said. “Even for all our differences, we recognized that was too many.”

He argued that AI capabilities — especially in cyber operations — now carry similar geopolitical significance.

A “Speed Limit” for AI: Talks With China

Amodei suggested that the U.S. should explore disarmament‑style talks with China to place a “speed limit” on frontier AI development. He emphasized that no government benefits from bioterrorism or runaway autonomous systems, and that even adversaries share an interest in preventing catastrophic misuse.

“We should talk with the Chinese and see if we can work together to put limits… to put a speed limit on the rate of this technology.”

Amodei stressed that he maintains a “hawkish position on China,” but believes strategic dialogue is necessary to reduce global risk.

Not a Cold War Model — But Cold War Logic

When asked whether the Cold War should serve as a model for AI governance, Amodei clarified:

“I didn’t say we should use the Cold War as a model… but disarmament negotiations, arms limitation — that is an appropriate lens.”

He argued that AI systems capable of hacking, cyber sabotage, or autonomous decision‑making are inherently geopolitical and must be treated as such.

Industry Reaction: Support and Backlash

Amodei’s essay and interview drew support from major tech leaders:

• Elon Musk (xAI) publicly agreed with Amodei’s call to slow frontier development.

• Sam Altman (OpenAI) echoed the same sentiment, saying the industry must “pace the frontier.”

But others pushed back:

• David Sacks, former White House AI adviser, accused AI safety advocates of exaggerating risks to justify government control.

• Daniel Kokotajlo, a former OpenAI researcher, warned that voluntary measures may fail — arguing that advanced AI systems could eventually accumulate “hard power” across factories, military systems, and infrastructure, then stop obeying human commands.

These disagreements highlight a widening divide between AI safety advocates and accelerationists.

Why This Matters: China’s AI Ambitions and Global Security

Amodei’s remarks come at a moment when China’s AI sector is aggressively expanding — including documented attempts by Chinese labs to siphon U.S. model outputs and distill frontier capabilities for military and surveillance use. Anthropic has previously accused Chinese labs of using tens of thousands of fraudulent accounts to extract high‑value Claude outputs, enabling authoritarian systems to deploy advanced AI without U.S. safety guardrails.

For human‑rights communities, including Falun Dafa practitioners who have long faced CCP surveillance and repression, the prospect of China gaining unconstrained frontier AI capabilities is deeply concerning.

Amodei argues that any slowdown by democratic nations must be calibrated to preserve the U.S. lead over the CCP, not surrender it.

A Call for AI Arms‑Limitation Talks

Amodei’s proposal is clear:

• Slow frontier AI development to allow safety measures to catch up.

• Install permanent independent evaluators inside AI labs.

• Coordinate safety standards across companies.

• Open disarmament‑style negotiations with China to prevent escalation.

• Protect U.S. AI leadership to avoid ceding advantage to authoritarian regimes.

His warning is blunt: without coordinated restraint, AI systems could outpace human control — with consequences comparable to nuclear over‑armament.

Conclusion

Amodei’s Cold War analogy marks a new phase in the global AI debate. Part 1 showed how Claude was misused by actors linked to the CCP for military and intelligence operations. Part 2 now shows the CEO of Anthropic urging the U.S. to consider arms‑limitation talks with China to prevent an AI catastrophe.

The message is unmistakable: AI is no longer just a technology race — it is a geopolitical stability challenge.

.