Theft. Rivalry. Restrictions. Espionage. And declarations. The US and China artificial intelligence race has all the trappings of a thriller – one that now appears to be coming to a head.
Released last month by Beijing-based Z.ai, AI model GLM-5.2 has publicly drawn admiration from Silicon Valley. The surrounding buzz is on the model’s ability to complete complex tasks with minimal prompts and rivals its US counterparts at a fraction of their cost.
Some have called it a “mini DeepSeek moment”, a reference to January 2025, when China released the chatbot that took the US-dominated AI industry and markets world-over by surprise. China, it turned out, could develop cutting edge AI without advanced Nvidia chips – on which US’s ChatGPT and Claude rely, and which have deliberately been kept out of Beijing’s reach.
Also read: Negotiating AI paradigms
China closing the gap?
GLM-5.2 currently ranks fifth on Artificial Analysis’ intelligence leaderboard and second on Code Arena’s front-end coding rankings, news agency Reuters reported.
Z.ai founder Tang Jie told xAI’s Elon Musk on social media that GLM-5.2 was on par with Claude Opus 4.8 and the goal was to reach capabilities of Claude Fable 5 – Anthropic’s most powerful model available to the public – by the first quarter of 2027.
The episode puts GLM-5.2 at the heart of a growing debate about whether China is finally catching up to the US in the AI race.
“We now have a Chinese open-weight model that is as good as the currently available models from OpenAI and Anthropic,” David Sacks, former AI czar in US President Donald Trump’s administration, said last week.
The timing of this debate isn’t incidental. GLM-5.2 was released the same month that saw Washington briefly impose curbs on Anthropic, along with other flashpoints, all of which have raised concerns about erratic US regulation of an AI industry keen to gallop ahead.
Also read: India must ride the next AI wave, not chase past ones
Mythos and Fable shutdown
In early June, Anthropic released Fable 5 — a public version of its most advanced model Mythos, which was not made available for wider use because it’s so powerful that it can easily detect vulnerabilities in software and aid bad actors in hacking and other illegal activities.
Three days later, the US government ordered the company to cut off access to both the models for anyone who was not an American citizen. Unable to easily verify users’ nationality, Anthropic shut the models down worldwide.
The Trump administration had said the trigger for the directive was a discovered “jailbreak” that could coax Fable 5 into producing dangerous information to users.
Washington partially eased the curb last week, and allowed Anthropic to restore access to Fable from 1 July.
For some observers, the brief shutdown was linked less to safety concerns than to a falling out between Anthropic and the Trump administration. In February, Anthropic had declined to let the Pentagon use its AI without limits on surveillance and weapons use, prompting Trump to order government agencies to stop using all products by the company and describing it as a firm of “Leftwing nut jobs” in a social media post.
Adding to this episode is speculation that the Trump administration, whose approval ratings are down as midterm elections are coming up, is leaning on regulating the AI industry that many in the US believe could hit their jobs and employment opportunities.
Also read: HCLTech signs massive $1.14 billion deal with Europe-based Fortune Global 50 firm to develop AI model
A ‘missing’ machine, and Anthropic accuses Alibaba
The same weeks saw separate flashpoints.
US commerce secretary Howard Lutnick told senior leaders at Dutch chip-equipment maker ASML that Washington suspected one of its extreme ultraviolet (EUV) lithography machines — used to manufacture the most advanced processors — may have reached China in violation of American export controls, Bloomberg reported.
ASML has never been permitted to ship the systems to China, and the company denied any breach, saying in an internal document that all 314 EUV machines in operation worldwide were accounted for outside China.
Senior US officials told Bloomberg they had evidence ASML had shipped specialty equipment linked to EUV systems to China, though they declined to make that evidence public. ASML’s access restrictions matter because the absence of EUV tools is the single toughest constraint on Huawei Technologies, China’s leading rival to Nvidia in AI chip design.
A third dispute erupted after Anthropic accused Alibaba Group of illegally accessing its AI model. The claim sent Alibaba’s Hong Kong-listed shares to a 16-month low, and shares in other Chinese AI developers, including Xiaomi and Baidu, also fell more than 3% on the news.
Bloomberg Intelligence analyst Robert Lea said the episode, combined with the Anthropic export curbs, signalled that “Chinese AI models face an elevated risk of a US ban”.
The race & its stakes
The AI contest has drawn comparisons with the 20th century nuclear race between the US and Soviet Union. But that had a clear endgame: the nuclear weapon. The AI one doesn’t, at least not yet.
For now, both the countries, experts say, have their distinct advantages.
Kyle Chan, research fellow at Brookings Institution who specialises in Chinese tech and industrial policy, said in a podcast last month that US’s strength was in the “virtual world” – advancing research, and developing AI models and the software and commercial ecosystem around them.
China’s forte, he said, was the “physical world” of manufacturing and supply chain innovations. There are few powers that can make strides in a field where they don’t have access to the best inputs, Chan said.
