ByteDance is pursuing legal action against a former intern, seeking 8 million yuan (~$1.1 million) in damages for allegedly sabotaging the company’s AI language model training infrastructure.

TikTok’s parent company filed the lawsuit against Tian Keyu in Beijing’s Haidian District People’s Court, Reuters cites state-owned media outlet Legal Weekly. Tian, identified by Chinese media as a postgraduate student at Peking University, is accused of deliberately manipulating code and making unauthorized modifications that disrupted the company’s model training tasks.

The case has attracted significant attention within China, mainly due to its focus on large language model (LLM) training technology amid growing global interest in generative AI developments. Reuters notes that while employee-employer litigation is common in China, legal action against an intern involving such substantial damages is unusual.

ByteDance, which reached a valuation of $300 billion last month, addressed rumors about the case’s impact in an October social media post, stating that claims of millions of dollars in losses and involvement of over 8,000 graphics processing units were “seriously exaggerated.” The company terminated Tian’s internship in August.

The case emerges as competition intensifies in China’s AI sector, with major technology companies investing heavily in developing and protecting their AI capabilities. 

According to the China Organization Data Service, cited by the Global Times, as of October 31 this year, the number of AI organizations in the country had reached nearly 1.9 million, with over 60,000 enterprises nationwide in the manufacturing sector implementing AI.

“China’s AI industry steadily ranks in the world’s first tier, and it is a global leader in terms of AI algorithms for language models of all sizes,” Pan Helin, a member of the Expert Committee for the Information and Communication Economy under the Ministry of Industry and Information Technology, told the Global Times.

Reportedly, ByteDance plans to use Huawei chips to build its AI models while its web crawler, Bytespider, scraps web data approximately 25 times faster than OpenAI’s GPTbot.