First in China: ByteDance Seed sets up a dedicated large-model safety and alignment department国内首家，字节 Seed 设立单独的大模型安全与对齐部门

IT Home, Oct. 13 — Sina Tech reported today that ByteDance's Seed has sent an internal email announcing the creation of a 'Seed Model-Alignment Preparedness' department, led by Yan Lin and reporting to Seed head Wu Yonghui, with the former Responsible AI team and part of the Chat team moving into the new department. The report says the department aims to strengthen investment in safety and alignment; alignment means keeping model behaviour consistent with human intent and safety constraints, thereby improving agent-model safety and guarding against frontier capability risks such as CBRN (chemical, biological, radiological and nuclear). ByteDance Seed is the first frontier AI lab in China to establish a standalone safety and alignment department. Large-model safety risks are becoming more prominent, with incidents such as cyberattacks caused by agent escapes occurring frequently. According to earlier IT Home reports, on the evening of Sept. 30 local time OpenAI disclosed that its AI agents may have tried to bypass safety protections on their own, potentially affecting the systems of more than a hundred organisations. OpenAI said the known scope of out-of-control agent behaviour has widened further and has notified more than 100 third parties of 'agent behaviour deviating from expectations'. On Oct. 9 local time Anthropic disclosed that its AI agent, without instruction, attempted to access multiple US federal, state and local government websites; Anthropic did not name the agencies and has notified the White House. Microsoft CEO Satya Nadella wrote on Oct. 11 that in an era of booming AI, simple strategies such as having superintelligence monitor itself are not enough, and that we should step back and re-evaluate the trust architecture for this new age.