Chinese technology giant ByteDance’s artificial intelligence (AI) research unit, Seed, has established a dedicated safety and alignment team and launched a global recruitment drive for 10 researchers, signalling a stronger focus on managing risks associated with increasingly capable AI models.
The TikTok-parent’s hiring push comes as concerns over AI-agent safety intensify in China and the rest of the world, particularly as developers build systems capable of independently accessing the internet and carrying out complex tasks.
In a recruitment announcement published on its official WeChat account on Wednesday, Seed said it planned to make a long-term investment in AI safety, arguing that alignment and preparedness would become increasingly important as model capabilities advance.
“AI alignment refers to efforts to ensure that AI systems behave in accordance with human objectives and safety requirements, including when they undertake complicated, multi-step tasks,” ByteDance stated further.
Seed has put safety as the core capability needed to deploy advanced AI systems at scale, instead of treating the concept as merely a precaution against potential harm.
The announcement marks a rare public disclosure by a major Chinese AI developer of a standalone team dedicated to frontier-model safety, alignment and preparedness.
State-backed newspaper The Paper reported on Tuesday that Seed was the first Chinese frontier AI laboratory to establish an independent department focused on safety and alignment.
The new team will operate in Beijing, Shanghai and San Jose, California. Seed said it would receive computing capacity and research resources comparable to those available to teams developing its frontier AI models.
Its work will span several areas, including safety training throughout model development, automated vulnerability testing and tools to monitor and explain AI behaviour.
Researchers will also develop testing environments for AI agents and investigate ways to prepare for emerging risks as systems become more capable and autonomous.
Seed said it intended to subject its work to independent testing and collaborate with universities and research institutions. Safety research would be incorporated across the model lifecycle, from training and evaluation to deployment.
ByteDance’s initiative comes amid broader US-China discussions about AI safeguards, potential safety incidents and channels for communicating about major risks.
Both countries are competing to advance increasingly sophisticated AI systems, while questions over how to test, monitor and control them are gaining prominence.