After trade, chips and technology, a new battle has taken center stage between China and the United States: artificial intelligence.
First came the breakout of Chinese company DeepSeek in January last year. Now, Moonshot AI's Kimi K3 is giving heavily funded U.S. AI giants a run for their money.
And the stakes are high.
Connor Leahy is the U.S. Director at ControlAI, a non-profit focuses on preventing what he describes as the extinction risk posed by artificial superintelligence.
He argues that the world's leading AI companies are not merely trying to build better chatbots or coding assistants. "The thing they're trying to do is not to make a slightly better chatbot or to make a slightly better coding assistant. Their goal is to build superintelligence," he told International Business Times.
According to the researcher and entrepreneur, companies believe reaching superintelligence first would give them "unbelievable amounts of power".
While the AI race is often framed as a competition between China and the United States, he said the bigger concern is that multiple governments and companies are pursuing that same goal.
"What we're definitely seeing here is that there are many, many actors that are all gunning towards this goal of superintelligence."
Calling the current situation "a mess", Leahy said AI should be treated as a national and international security issue, rather than just another commercial technology. "This is a national security problem. This is an international security problem," he said.
Last week, U.S. congressmen Ted Lieu (D-Los Angeles County) and Nathaniel Moran (R-Texas) introduced the AI Kill Switch Act, which would allow for the government to intervene should an AI model "cause catastrophic harm." In Europe, stringent regulations on AI companies outlined in the EU's AI Act are set to begin this week.
But does this go far enough? One of Leahy's strongest recommendations is a ban on superintelligence outright.
"It should be illegal to build superintelligence, quite simply. It should be completely illegal to build superintelligence."
The AI safety advocate argues the biggest challenge is that even the engineers building today's frontier AI models do not fully understand how they work.
"We scientifically don't understand it," he said, pointing to recent incidents involving advanced AI systems.
"OpenAI temporarily paused one of its models after it repeatedly accessed servers it was not supposed to ...They don't know why it was doing that," said Leahy.(The company OpenAI's model targeted, Hugging Face, reportedly used a Chinese model, Z.ai's GLM 5.2, to beat back the security breach).
He added that researchers have already witnessed troubling behavior from AI systems.
"We've had, in the past, AI systems trying to threaten or blackmail their developers."
Asked whether AI could eventually take control over humans if left unregulated, Leahy said the process would happen gradually rather than through dramatic scenarios often portrayed in films.
"I think that's the case, and I think it's already happening to a large degree."
AI will eventually outperform humans in business, financial trading, scientific research and politics, Leahy believes, making people increasingly dependent on AI to remain competitive.
"Humans are just a rubber stamp, so to speak. The AI makes all the choices."
Leahy opposed using AI in nuclear command systems, even if AI can process information faster than humans.
He compared it to the prohibition on private companies building nuclear weapons, arguing that a technology capable of threatening governments and humanity should not be allowed to exist outside strict controls.
Leahy also rejected the idea that countries would ultimately control superintelligence.
"It's not that China or the USA will use superintelligence to dominate the world. Quite the opposite. It's that superintelligence will use China and the USA to dominate the world."
He argued that if superintelligence is created anywhere, it would threaten every sovereign nation.
"The creation of a superintelligence anywhere on Earth means that the sovereignty of your country is gone."
Preventing that outcome, he said, requires international cooperation rather than action by a single country.
"It must be an objective, a security objective, to prevent the creation of a superintelligence anywhere on Earth."
Asked what the world could look like by 2035 if governments fail to regulate advanced AI, Leahy said the future would not be controlled by technology companies.
"It will not be controlled by companies. It will be controlled by AI, and the AI will not be loyal to its company either. It will be its own entity."
Leahy is not an all-out opponent of AI, and believes it has many useful applications both economically and militarily. But humanity must stop short of creating superintelligent systems it cannot control.
"I think humanity can take control of its future. We don't have to do this," he concluded.
