According to an internal email sent recently, ByteDance Seed announced the establishment of the "Seed Model-Alignment&Preparedness" department, led by Yan Lin, who will report to Seed head Wu Yonghui. The original Responsible AI team and some members of the Chat team will be transferred to the new department.

The establishment of this department aims to strengthen investment in safety and alignment. The core of alignment is to ensure that model behavior remains consistent with human intent and safety constraints, thereby enhancing the safety capabilities of Agent Models and preventing risks from frontier capabilities such as CBRN (Chemical, Biological, Radiological, and Nuclear).

Safety risks associated with large models are becoming increasingly prominent, with incidents such as cyberattacks triggered by agent escape occurring frequently. Previously, the safety and alignment work of domestic large model research teams was mostly incorporated into the company's overall safety and compliance system. ByteDance Seed is the first frontier AI laboratory in China to establish a separate safety and alignment department.

Internally, the Seed team judges that as model capabilities grow stronger and application scenarios become more complex, safety and alignment are no longer just about risk prevention but have become core capabilities that determine whether models can be applied at scale.

It is understood that the Seed safety and alignment department will continue to invest in areas such as model specification formulation, safety evaluation system construction, alignment training, protection system development, and risk monitoring. It will also strengthen public cooperation with external research institutions and third-party evaluations. Ultimately, the goal is to enable frontier models to effectively complete tasks within safety and authorization constraints. As model capabilities and autonomy increase, the department aims to continuously reduce risks of misuse and behavioral deviation, and establish a comprehensive safety system to address unknown risks.