ByteDance Seed Announces AI Safety and Alignment Department

THE ESSENTIALS

The reported reorganization gives safety and alignment work a dedicated department within Seed.

A dedicated team

ByteDance’s Seed lab has announced a new Model-Alignment & Preparedness department in an internal email, The Paper reported on October 13. Yan Lin will lead it and report to Seed head Wu Yonghui. Staff from Responsible AI and part of the Chat team will move into the unit.

Agent safeguards

Its planned work includes behavior specifications, safety testing, alignment training, protective systems and risk monitoring. The report says Seed also intends to cooperate publicly with external research institutions and third-party evaluators. These are organizational plans, not evidence that the associated model risks have been resolved.