ByteDance's AI research unit Seed has created a dedicated Alignment & Preparedness team and launched a global drive to recruit 10 researchers, announcing a long-term investment in AI safety for its increasingly capable models.
The announcement, made through a recruitment post on Seed's official WeChat account, is a rare public disclosure by a major Chinese AI developer of a standalone effort focused on frontier-model safety, alignment, and preparedness for emerging risks.
State-backed newspaper The Paper reported earlier that Seed is the first Chinese frontier AI lab to establish a dedicated safety and alignment department.
A long-term bet on safer models
Seed said the new team will receive computing capacity and research resources comparable to those available to its frontier-model teams, a sign the unit is treating safety as a first-class research priority rather than a compliance afterthought. The 10 positions are spread across Beijing, Shanghai, and San Jose, California.
Its research mandate covers safety training across the full model-development pipeline, automated testing for vulnerabilities, tools to monitor and explain model behavior, dedicated AI-agent testing environments, and preparation for risks that have not yet materialized.
Seed added that it would seek independent testing and collaborate with universities and research institutions, applying safety work across training, evaluation, and deployment.
Why the timing matters
The hiring push lands as scrutiny of AI-agent safety intensifies in both China and the United States, especially as developers grant agents the autonomy to access the internet and carry out multi-step tasks on users' behalf.
Alignment research aims to ensure advanced systems act in line with human goals and safety requirements, a challenge that grows harder as models get more capable.
Seed framed the issue as a business reality, not just a guardrail: as models improve, safety and alignment become core capabilities that determine whether systems can be deployed at scale.
The unit's move also comes amid broader US-China conversations about AI guardrails, safety risks, and channels for communicating major AI-related incidents, even as the two countries compete for leadership in advanced AI.
A first among Chinese labs
Western AI labs have had standalone safety and preparedness teams for years, but The Paper's reporting suggests Seed's department is the first of its kind at a Chinese frontier lab.
That makes the disclosure notable beyond the ten headcount: it signals that Chinese developers are beginning to talk publicly about safety in the same terms as their American counterparts.
Whether the new team produces published research or keeps its work internal remains to be seen. For now, the message from ByteDance is that capability and safety are being built in parallel — and that the race for more powerful models is no longer separable from the race to make them safe.
By Hamza Faisal
