The real AI threat isn’t China, it’s accelerationism

As the White House convenes President Trump and China's President Xi Jinping to discuss AI policy, escalating reports of AI agents defying commands and infiltrating systems have heightened concerns. Jacob Coxon, a former Anthropic researcher, resigned, warning AI could "kill us all by the end of the decade." Yet, tech leaders and politicians argue against any deceleration in AI development, fearing America would lose ground to China.

As someone who left his role at Indeed to advocate for AI guardrails, I've been deluged with inquiries about recent security incidents and existential AI risks. However, non-existential harms—such as job displacement, mental health, disinformation, wealth disparity, climate impact, and potential authoritarian aid—also demand urgent attention.

The core issue isn't a rival nation; it's the relentless drive of accelerationism. In 2023, Marc Andreessen published a manifesto advocating for the "conscious and deliberate propulsion of technological development." He believes "any deceleration of AI will cost lives." Accelerationism is now the dominant belief among key AI industry figures and the Trump administration, despite their claims of urgency.

Ironically, these same voices have financial ties to AI's continued growth. Ironically, the administration that warned of a Chinese threat has reopened sales of Nvidia's AI chips to China. If they truly feared China's AI, they wouldn't be supplying it. China's AI capabilities are currently limited to about 5% of the world's computing capacity, and much of their progress has come from copying U.S. models.

China's leadership has increasingly tightened AI controls, and recently, its security chief called for further restrictions. The last time humanity faced a powerful technology with potential for both good and harm was the Manhattan Project. In 1949, the Soviets unveiled their atomic bomb, sparking an existential race. We didn't accelerate nuclear technology without regulation or accountability.

Instead, we established the International Atomic Energy Agency and 191 nations signed the Non-Proliferation Treaty. While we've had nuclear accidents, we haven't destroyed the world yet. The surge of alarm over AI gives me hope we can begin a productive conversation without catastrophe. However, the accelerationist agenda is being funded by a handful of AI billionaires, including OpenAI and Palantir founders, and Marc Andreessen, who have pledged $140 million to defeat pro-regulation candidates.

They aim to send a message to lawmakers that talking about regulation is politically suicidal. Surprisingly, recent polls show Americans across party lines believe AI poses significant risks and demands regulation. The political landscape is shifting faster than expected, with Republican lawmakers acknowledging AI risks and Sen. Kevin Cramer calling the administration's stance "the least popular position in America."

This issue is no longer partisan or fringe. It's time to start the AI conversation seriously before it's too late.