ByteDance’s AI research division, Seed, has established a new department focused on safety, alignment, and risk management. Known as the Model-Alignment & Preparedness department, this unit is the first of its kind to operate as a standalone entity within a Chinese AI laboratory.
The department is led by Yan Lin, who reports directly to the head of Seed, Wu Yonghui. The company has initiated a global recruitment drive to hire an initial group of ten researchers for the team. Positions are currently open in Beijing, Shanghai, and San Jose.
The new unit will focus on several core areas, including defining model behavior specifications, conducting safety evaluations, and implementing automated red-teaming. Additionally, the team will address high-risk capabilities related to chemical, biological, radiological, and nuclear threats.
This restructuring elevates safety functions to the same organizational level as model capability research. Previously, such efforts at Seed and other Chinese labs were typically integrated into broader departments focused on content moderation and compliance.
