Are Google and OpenAI using technicalities and loopholes to circumvent US export restrictions to China?
Are Google and OpenAI using technicalities and loopholes to circumvent US export restrictions to China?

Google highlights 'strong protections' amid AI sales to Chinese subsidiaries

Cody Combs

Some tech firms are reportedly using loopholes in US law to sell their products to Chinese companies the Pentagon has deemed worrisome.

According to the Financial Times, Alphabet-owned Google and OpenAI - the maker of ChatGPT - have been selling their advanced AI technology to Chinese tech giants blacklisted by the Defence Department.

The sales are reportedly being made to various "Singapore-based subsidiaries" of China's Alibaba, Tencent and Baidu.

Despite the Pentagon labelling all three companies as being "indirectly affiliated" with China, making the sales in Singapore to subsidiaries of the companies skirts US restrictions.

As the global race for AI dominance has heated up in recent years, the US and China have emerged as the top competitors, with Washington slightly leading but Beijing catching up.

A Google representative told The National that the technology company was committed to complying with all applicable export control and trade laws.

They added that the company has "strong terms of use and protections", including disabling accounts and blocking access to AI services.

Founded in 1998, Tencent enjoyed uninterrupted growth from when it went public in 2004 until this year. Reuters
Founded in 1998, Tencent enjoyed uninterrupted growth from when it went public in 2004 until this year. Reuters

Another Google representative said that the company differed from other firms in that it doesn't operate any cloud computing regions or data centres in mainland China.

The representative also said that while the company's AI services are available in international commercial hubs of Hong Kong and Singapore, Google's "rigorous acceptable use" and "security policies" still apply.

One of the main issues concerning US technology experts, executives and elected officials is that China is widely speculated to have been stealing US proprietary information behind powerful US AI models through a process known as distillation.

Executives for both OpenAI and Anthropic have expressed concern about distillation in recent months.

A Google representative told The National that while the company takes concerns about distillation very seriously, simply "restricting sales to certain geographies" doesn't lessen the likelihood that distillation will occur, and that Google routinely tracks those attempts and shares them with other companies.

Collectively, the Bats - Baidu, Alibaba and Tencent - have lost around $165 billion in value year-to-date. Photo: Reuters
Collectively, the Bats - Baidu, Alibaba and Tencent - have lost around $165 billion in value year-to-date. Photo: Reuters

"We’ve known for some time that Chinese subsidiaries outside of China access US chips under export control in servers also located outside of China, and there has been a movement in Congress to close this loophole by extending export controls to these remote data centers," said Mark MacCarthy, a senior fellow at the Institute for Technology Law and Policy at Georgetown University in Washington.

He also pointed out an ongoing push from some lawmakers to prevent Chinese AI models like DeepSeek from being used in the US.

Mr MacCarthy described those efforts as "protectionist measures aimed at preserving the US AI market for US AI companies."

Alibaba, Baidu and Tencent have not yet responded to The National's requests for comment.

OpenAI has also not responded to requests for comment. According to the FT, the company said it recently suspended "Alibaba-affiliated users’ access to its API ... over concerns about illicit use."

Updated: July 10, 2026, 7:38 PM