Outside AI blocked by confidentiality category
China's ByteDance will, from November, keep outside AI from accessing confidential internal information, TokenPost reported on Nov. 3. The restriction covers code repositories classified as "L4" in its internal confidentiality categories and documents in "Feishu" (known overseas as Lark), the document service used across the company. Anthropic's Claude and OpenAI's Codex are named as outside tools subject to the restriction.
@LeroyLi311063, an X account that posts information about China, reported the same content on the same day. It did not, however, cite a source.
Personal AI accounts also banned
Employees will also be banned from accessing restricted information through AI accounts they bought personally. Some situations in the company's own AI unit, "Seed," are reportedly exempt.
Hong Dingkun, ByteDance's vice president in charge of technology, reportedly said that a leak of content at L3 or above could have serious consequences. He also reportedly said there was no need to ban other companies' AI tools to spread the company's own AI, and that forcing employees to use it would have little effect.
The TokenPost article does not cite an official ByteDance announcement or internal documents. The scope and timing of the restriction remain at the level of a report.
Cursor restricted in 2025 too
ByteDance also restricted employees' use of the AI coding tools Cursor and Windsurf in May 2025, saying work data had been stored in personal accounts.
