SINGAPORE – ByteDance, the parent company of TikTok, is suing a former intern for 8 million yuan (US$1.1 million), accusing him of deliberately sabotaging its AI large language model (LLM) training infrastructure. The case has captured significant attention in China, given the ongoing AI arms race and the unprecedented nature of the allegations.

The lawsuit, filed in the Haidian District People’s Court in Beijing, centres on allegations against Tian Keyu, a postgraduate student at Peking University, as reported by the state-owned Legal Weekly.

Industry observers noted that disputes between companies and employees are common in China, but a legal action targeting an intern for such a substantial sum is highly unusual, raising questions within the tech and legal communities, Reuters reported.

The focus on LLM training – a critical component of generative AI capable of producing text, images, and other outputs – further underscores the high stakes of the case. ByteDance’s claim points to deliberate acts of sabotage that allegedly disrupted its research processes and wasted resources.

Details of ByteDance's allegations

According to ByteDance, Tian manipulated code and introduced unauthorised modifications between June and July 2024. His actions reportedly disrupted AI training tasks, wasted resources, and undermined the team’s progress. An internal investigation, cited in various media reports, revealed that Tian had:

• Implanted a backdoor in checkpoint models using unsafe serialisation formats and granting unauthorised access
• Inserted random delays into training processes, terminated sessions, and reversed training steps to derail progress
• Altered colleagues’ model weights and attended troubleshooting meetings under false pretences, using the insights to refine his sabotage tactics

The company claims Tian’s actions were uncovered after disruptions in AI model training prompted an internal review.

ByteDance subsequently terminated Tian’s internship in August and reported the incident to his university, citing his repeated denials of wrongdoing during the investigation.

Also Read: Quit-Tok: Gen Z trend exposing toxic workplaces

ByteDance’s response and security concerns

While the lawsuit seeks significant damages, ByteDance has sought to downplay the broader impact of the incident. In an October social media statement, the company dismissed rumours that the sabotage had cost millions of dollars or involved over 8,000 graphics processing units, calling such claims “seriously exaggerated.”

ByteDance said the disruptions were limited to a specific research project and did not affect its core commercial operations. However, the incident has drawn attention to vulnerabilities in AI development and the need for robust internal security protocols to prevent insider threats.

Also Read: TikTok announces job cuts amid widespread tech layoffs

Tian’s defence and reputation management

Tian has denied the allegations, maintaining that he did not cause significant harm to ByteDance’s operations. Reports suggest he has been actively attempting to salvage his reputation, disputing the severity of the claims.

However, details of his defence have not been widely reported, leaving questions about the veracity of ByteDance’s accusations and the extent of any actual damage.

AI industry implications

The case highlights a critical issue in the fast-evolving AI sector: safeguarding intellectual property and proprietary systems.

As companies like ByteDance race to develop cutting-edge AI technologies, they face increasing risks of insider threats, espionage, and sabotage. This lawsuit underscores the necessity of stringent security measures, even at the intern level, in protecting valuable R&D initiatives.

While ByteDance’s reputation as a global tech leader remains intact, the high-profile nature of the case serves as a cautionary tale for businesses navigating the complexities of innovation and internal security in the AI industry.