(Yicai) Oct. 13 -- TikTok owner ByteDance's artificial intelligence research team Seed plans to establish a dedicated department to strengthen AI safety and alignment as models become increasingly capable and autonomous.

Seed will be the first frontier AI laboratory in China to establish such a department, The Paper reported today. Named Seed Model-Alignment & Preparedness, the new unit will focus on ensuring that AI models behave in accordance with human intentions and safety constraints.

The department will step up investment in improving the safety of AI agent models and preventing risks associated with advanced chemical, biological, radiological, and nuclear capabilities, according to the report.

It will also seek to reduce the risks of misuse and unintended model behavior as AI capabilities and autonomy advance, while building a comprehensive safety framework to address emerging and unforeseen threats.

Citing an internal company email, The Paper reported that the department will continue investing in model specifications, safety assessment systems, alignment training, protective mechanisms, and risk monitoring. It will also strengthen open collaboration with external research institutions and third-party evaluators.

The ultimate goal is to ensure that advanced AI models can effectively carry out tasks while operating within safety and authorization constraints, the email said.

Founded in 2023, Seed researches multimodal large language models, generative media, AI for science, and robotics, with the aim of advancing artificial general intelligence, according to ByteDance's website. The team has laboratories and job openings in China, Singapore, and the United States.

Editor: Emmi Laine