US-China AI Talks Begin Amid New $1 Trillion 'Cold War'

A new US-China dialogue on artificial intelligence governance masks a deepening technological rivalry that one top AI firm is calling a new Cold War.

The United States and China have initiated a formal dialogue on artificial intelligence governance, a move aimed at managing risks even as leading US firms like Anthropic frame the competition as a zero-sum "Cold War" for democratic survival, with potentially trillion-dollar valuations hanging in the balance.

"The two heads of state held constructive discussions on AI-related issues and agreed to launch an intergovernmental dialogue on artificial intelligence," Chinese foreign ministry spokesman Guo Jiakun said at a May 19 press briefing, confirming earlier remarks by US Treasury Secretary Scott Bessent.

The dialogue follows a period of escalating tension, with the White House accusing Chinese entities of "industrial-scale" tech theft and Beijing blocking a key acquisition by Meta. Concurrently, US AI firm Anthropic withheld its powerful "Mythos" model from public release to prevent exploitation, highlighting cybersecurity fears shared by both nations.

At stake is leadership in a technology sector poised for explosive growth, with Anthropic alone potentially eyeing a trillion-dollar valuation. The central debate is whether cooperation on safety, particularly preventing advanced models from falling into non-state actors' hands, can coexist with a fierce battle for technological supremacy that could define the 21st-century global economy.

Anthropic's War Footing

While diplomats talk, leading AI developers are choosing sides. Anthropic, a key competitor to OpenAI, has positioned itself as a 21st-century anti-China think tank, according to a Wall Street Journal report from May 20, 2026. CEO Dario Amodei has been explicit in claiming the survival of democracy depends on the US winning the AI race against China.

In a move reminiscent of the original Cold War's intellectual infrastructure, the company recently published a 5,500-word manifesto on how to conduct this new AI Cold War. This strategic pivot raises questions for the company, which could pursue a trillion-dollar IPO later this year: is it a business or a national security asset? The last time a similar dynamic emerged, the RAND Corporation was spun off from Douglas Aircraft to become America's premier Cold War think tank, allowing the parent company to focus on its commercial and defense business.

Anthropic's immediate concern is the uncannily effective ability of its latest models, like Mythos, to find critical software vulnerabilities. The company and others in the US believe Chinese state and criminal hackers are at most a year behind in developing similar capabilities.

The Open-Weight Dilemma

A sinister hue is increasingly attached to Beijing’s preference for “open weight” AI models, which allow any user to adapt and modify them at will. Unlike their Western counterparts, Chinese AI builders typically don't reserve the ability to monitor usage or exercise a "kill switch" over antisocial activities. This raises fears in Washington that Beijing plans to put powerful cyber tools into the hands of anti-Western hackers while maintaining deniability.

The strategy also poses a direct threat to the business models of firms like Anthropic and OpenAI, which have invested billions in developing proprietary systems. Anthropic's manifesto identifies two key battlegrounds: restricting China's access to top-tier semiconductor chips and preventing "distillation"—the illicit use of Western models to train Chinese versions, effectively hijacking billions in US investment.

The recent dialogue, focused on creating a "protocol" to keep models from non-state actors, represents a narrow path for cooperation. It echoes a 2024 agreement between the two nations' leaders that humans must remain in control of nuclear weapons. Yet, with the immense economic and strategic power of AI at stake, the underlying conflict between an open, cooperative system and a closed, competitive one is only just beginning.

This article is for informational purposes only and does not constitute investment advice.