Beijing appears to be demanding that Chinese AI models have 'socialist values'

Beijing's aggressive ideological requirements regarding its homegrown LLMs could hold China's AI sector back, one expert said.

• China's government appears to be extending its AI crackdown.
• The CCP is requiring tech companies to submit AI models for review, The Financial Times reported.
• The intense scrutiny could slow China's AI development and commercialization.

The Chinese government appears to be extending its crackdown on AI.

Beijing has always had an intense censorship and online surveillance policy, which now extends to local AI development. The Chinese Communist Party has introduced several regulative measures to keep AI models in check and ensure that chatbots tow the party line.

The Financial Times reported that the latest development involves forcing China-based tech companies to participate in a government review of AI models.

Officials are testing the models to check they "embody core socialist values," per the report.

The CCP already has the final sign-off on all Chinese large language models (LLMs) and the data they are trained on. International models such as ChatGPT have also been blocked in China.

This means blocking sensitive keywords or information the CCP would rather not have in the public domain. Similar to the country's Great Firewall, Chinese chatbots already dodge questions about the 1989 Tiananmen Square massacre and comparisons between Xi Jinping and Winnie the Pooh.

Daniel Colson, executive director of the AI Policy Institute, said that China's aggressive ideological requirements regarding their LLMs could hobble the country's AI industry.

"Some policymakers can often incorrectly believe that the CCP is singularly focused on beating the US when it comes to AI development, but the fact of that matter is that their goals are more complicated than that. They have their own ideological interests conflicting with the goal of moving as fast as possible," Colson said.

China's leader, Xi Jinping, sees technology including AI as key to boosting the country's economy. While censorship requirements may slow China's AI development and commercialization, it remains a formidable force in the AI race.

Colson said some of the most advanced models in China are actually provided by US companies.

"Providing the CCP access to frontier models and their weights allows them to catch up faster," he said. "When the CCP has access to raw model weights, it makes it easier for them to create fine-tuned versions that allow such censorship."

China's plan for AI is about shaping reality and enforcing its power, according to internal documents previously reported by BI.

The US government has been trying to undercut China's AI efforts for some time, restricting access to powerful AI chips to slow the country's development.

"So much of the narrative in the US capabilities races is 'We have to stay relevant to outcompete China,'" Anna Yelizarova, a project lead at The Future of Life Institute, told BI.

But regardless of whether America's attempt to undercut China succeeds, the nation's conflicting visions for AI may already be playing to different audiences — and potential customers.

Several regimes worldwide would rather employ China's brand of highly regulated AI — something that could prove lucrative Beijing even if its tight control leaves it lagging behind the US in the AI race.