ByteDance miserably fired an executive for introducing malicious code into an AI model

ByteDance, the owner of TikTok, confirmed the dismissal of an intern for maliciously introducing malicious code into artificial intelligence models. The young man’s actions allegedly cost the company tens of millions of dollars in losses.

Image source: Claudio Schwarz/Unsplash

ByteDance, although it confirmed the dismissal of the intern, however, as Ars Technica reports, it denied rumors about multimillion-dollar losses and the lost computing power of 8,000 graphics processors, trying to stop conversations on this topic on Chinese social networks that are harmful to the company’s image. ByteDance called the claims “grossly exaggerated” and stressed that the damage was significantly less.

According to the official statement, the intern held a position on a commercial technology team and “maliciously interfered with model training tasks” as part of a research project. The company also accused the intern of adding false information to his social media profile indicating that his work was with ByteDance’s artificial intelligence lab rather than with the commercial technology team. ByteDance notified the intern’s university and industry associations about the incident, presumably to prevent him from trying to mislead others.

The company emphasizes that none of the commercial areas were affected and large language model (LLM) algorithms were also not affected. Despite this statement, some social media users continue to doubt the official version. Some claim that the commercialization team the intern worked on was previously part of an AI lab, and that the intern may have used “malicious code” to “deliberately sabotage model training for at least several months.” Other commentators suggest that ByteDance is downplaying the damage, and that “the loss of computing power alone could amount to several millions of dollars, if not tens, not to mention the loss of time, energy and wages of the entire research team.”