ByteDance Establishes Dedicated AI Safety Unit

ByteDance’s Seed AI lab has established a standalone Model-Alignment & Preparedness department, a move described by reports as the first such unit at a Chinese frontier AI lab. Led by Yan Lin, who previously oversaw alignment work for the Doubao model, the team reports directly to Seed head Wu Yonghui and incorporates ByteDance’s existing Responsible AI team and some Chat team staff. Its work will include model specifications, safety evaluations and training, monitoring and protections for AI agents, and preparation for emerging risks, including chemical, biological, radiological and nuclear threats. Seed has also announced a global recruitment drive for 10 researchers and plans to collaborate with outside researchers and independent evaluators, framing safety as a core capability for deploying increasingly capable models.