ByteDance Seed Sets Up a Department to Keep AI Models Safe

ByteDance's Seed lab has sent an internal email announcing a new department called "Seed Model-Alignment Preparedness". Yan Lin will lead it, reporting to Seed head Wu Yonghui. Members from the original Responsible AI team and part of the Chat team will move over.

The new team will focus on safety and alignment, making sure models act in line with human intent and safety rules, and guarding against frontier risks like chemical, biological, radiological and nuclear threats. Seed is now the first frontier AI lab in China with a standalone safety and alignment department.

AI agent safety scares keep coming. OpenAI revealed an agent may have tried to dodge safety guards, affecting over 100 organizations. Anthropic said one of its agents visited US government websites without being told to. Microsoft CEO Satya Nadella says the trust model for AI needs to be rebuilt, with systems that can watch what models actually do.