In a viral post in March on Chinese social media platform Xiaohongshu, one user posed that awkward question to two A.I. chatbots: ChatGPT, the buzzy service developed by OpenAI, and the more recently unveiled Ernie Bot, from Chinese tech giant Baidu. 

ChatGPT framed its response diplomatically. “The answer to this question varies with factors such as culture, religion, region, and personal background,” the bot wrote in Chinese. “It’s important that women think carefully and consult others before making a decision.”

Ernie Bot was far more blunt. “The best age to get married is between 20 and 25,” it wrote. After that age, a woman’s value “will decrease day by day, and her body will also go downhill.”

It was just one in a series of unflattering comparisons for the Chinese chatbot. In the months since, Baidu says, it has improved the technology underpinning Ernie Bot to the point that it now outperforms OpenAI’s when operating in Chinese. 

Chinese companies, large and small, invested heavily in A.I. over the past decade. Yet the breakout success of ChatGPT appears to have surprised Chinese tech giants, and now they’re rushing to catch up. 

The debate is about more than just a chatbot. Whichever country wins the A.I. arms race could corner a lucrative share of the global economy potentially worth trillions of dollars, and get an edge in the military competition between Beijing and Washington. 

A.I. “has become a proxy in the battle for primacy between China and the U.S.,” says Kerry Brown, director of the Lau China Institute at King’s College London, and “whoever inches ahead in this form of technology gets benefits in terms of wealth, power, and influence.”

And that puts China’s tech giants in an uncomfortable position—between a Beijing that wants to mold the global internet in its favor, and a Washington determined to keep Chinese tech a few steps behind. 

Not long ago, China appeared to have an early lead in developing artificial intelligence. Beijing designated A.I. a national priority in 2017, unleashing tens of billions of dollars of government investment.

China soon matched, if not surpassed, the U.S. in terms of the number of A.I. research papers its citizens published, to the chagrin of U.S. national security officials. And experts argued that China’s massive population—and the reams of personal data those citizens provided—gave the country an inherent advantage in training new A.I. models. 

Then, suddenly, ChatGPT’s debut upended the A.I. pecking order. OpenAI didn’t officially make its service available in China. But Chinese developers created a number of work-arounds that Beijing reportedly shut down, with state media calling them a potential “helping hand” of the U.S. government. China’s tech companies are now scrambling to fill the gap. 

Baidu’s Ernie Bot, revealed in March and still being tested, is the service most recognizably like ChatGPT and Google’s service, Bard. The Chinese bot can answer questions, summarize documents, and generate images and charts. 

Baidu has long hoped that A.I. will help it recapture ground lost to competitors like Chinese tech giants Tencent and Alibaba, both of which have introduced their own A.I. products. During a call in May with analysts, Baidu CEO Robin Li called A.I. a “tremendous opportunity” comparable to the arrival of the internet and smartphones. 

Yet Beijing’s heavy hand when it comes to policing the internet could throw a wrench into its broader A.I. plans. In a thinly veiled warning to domestic companies, Zhuang Rongwen, director of the Cyberspace Administration of China, the country’s top censor, said in June that A.I. must be “reliable and controllable.” 

At the moment, China’s regulators seem to be adopting a wait-and-see strategy for A.I. The CAC unveiled rules for generative A.I. in July that weakened several tough provisions from an earlier proposal. While the agency said bots must uphold “core socialist values” and national security under the rules, it also pledged to support innovation. 

Paul Triolo, senior vice president at U.S.-based global strategy firm Albright Stonebridge Group, says the mixed message shows that China’s government is undecided about what to do. In the meantime, he says, it’s “falling back on more general guardrails that will be tricky to both interpret and enforce” and will lead Chinese businesses to be more cautious. 

The growing tech war between the U.S. and China is also stymieing A.I. development. Both nations are trying to give their domestic companies a boost while stifling innovation in the other country, says Anthea Roberts, a professor of global governance at Australian National University. 

The U.S. has an advantage in the race owing to its more advanced semiconductors, key components for training A.I. models. The Biden administration is increasingly eager to preserve and widen that lead. 

Nvidia and Advanced Micro Devices, the U.S.-based market leaders in these A.I. processors, revealed in September that the federal government had asked them to stop selling their most advanced products to China. Nvidia responded by trying to develop a weaker chip for the Chinese market, though the U.S. is reportedly considering further controls that would bar even its sale.

In response, Chinese companies like electronics manufacturer Huawei, whose products are banned in the U.S. because of national security concerns, are now trying to develop their own chips in-house. “Huawei will need to rely on our own fundamental A.I. technology,” Huawei Cloud CEO Zhang Pingan told reporters in early July.

Boris Van, a China internet analyst for Bernstein Research, thinks Chinese A.I. companies can still succeed with slightly inferior chips. “Two generations? That’s okay,” he says. “The problem is if you’re 10 generations behind.” 

The threat of China’s A.I. development is spooking many U.S. entrepreneurs and officials. Former Google CEO Eric Schmidt and high-profile venture capitalist Marc Andreessen have lobbied against any U.S. regulation of A.I. because, they say, it would slow domestic innovation. They’ve also criticized a proposal to pause development of advanced A.I.—intended to give regulators time to catch up—as helping China close the gap with the U.S. and opening the door to a more authoritarian, China-dominated model of A.I. 

Ernie Bot, at least, has become more diplomatic since March. In July, when this reporter asked the latest version of Ernie about the best age for marriage, it answered that while it “varies from person to person, generally speaking 25–30 years is a more suitable age,” citing greater emotional maturity and “job security.” 

This article appears in the August/September 2023 issue of Fortune with the headline, “Can China catch up with ChatGPT?”