ByteDance, the parent company of TikTok, has fired a doctoral student intern following allegations of tampering with the company’s internal artificial intelligence (AI) model. The intern allegedly exploited a vulnerability on the AI development platform Hugging Face to access the model, which they reportedly used to disrupt training processes over nearly two months.
An internal investigation by ByteDance uncovered that the intern, a member of the company’s tech commercialization team, had become frustrated with ByteDance’s resource allocation for AI projects. The individual allegedly retaliated by gaining unauthorized access to the company’s language model infrastructure. The unauthorized access led to delays in model training and development, though ByteDance clarified that no online operations or active commercial projects were affected.
Local media outlet Jiemian reported that ByteDance’s automated machine learning (AML) team encountered difficulties in pinpointing the source of the disruption. According to sources close to the incident, the intern’s alleged activities hampered internal model training but did not affect the AI Lab directly.
ByteDance fired the intern in August after discovering the security breach. Rumors initially circulated that the incident caused millions of dollars in damages and affected over 8,000 GPU cards, claims that ByteDance has since dismissed as exaggerated. The company said that ByteDance’s Doubao commercial model and other core projects remained unaffected.