ByteDance curbs employee use of external AI tools like Claude, Codex

On 2026-11-04 at 16:59, TMTPost reported that ByteDance is tightening access to external AI tools, restricting products such as Claude and Codex and moving toward a self-developed, closed-loop setup in order to balance security and efficiency. The report did not specify the exact scope of the restrictions or when they would take effect.

One minute later, at 17:00, Landian reported that ByteDance is limiting employees' use of external AI tools to access internal materials, with Claude Code and Codex among the tools involved. The restricted materials include internal Level 4 code repositories and Feishu documents, and even AI accounts that employees purchase individually cannot access them. ByteDance's vice president of technology responded on the internal network, saying that the consequences of a leak of Level 3 or higher content are very serious and that using these external AI tools could lead to sensitive materials being exposed.

The two reports differ slightly on the scope of the restrictions: TMTPost refers only to products such as Claude and Codex in general terms, while Landian names Claude Code and Codex specifically and identifies the affected material levels. Neither report states when the restrictions take effect or exactly how they are enforced. As of the latest reporting, ByteDance has tightened access to external AI tools and its technology leadership has commented on the internal network; what happens next is not covered in either report.