TikTok owner ByteDance has established a dedicated AI safety team and launched a global recruitment drive for 10 researchers, promising resources comparable to those available to teams developing its most advanced AI models.
Its Seed research unit has formed an Alignment & Preparedness team to strengthen safeguards, test potential vulnerabilities and help ensure its AI models behave in line with human intentions and safety requirements. The move comes as technology companies face growing scrutiny over how increasingly capable AI systems and autonomous agents are tested and controlled.
AI alignment refers to efforts to ensure advanced AI systems act in line with human goals and safety requirements, particularly when carrying out complex, multi-step tasks. ByteDance says it intends to invest in this work over the long term as its models become more capable.
Growing Concerns Over AI Safety
AI agents can perform multi-step tasks with varying degrees of autonomy, including interacting with online tools. As these capabilities expand, developers face the challenge of identifying vulnerabilities and preventing systems from behaving in unintended ways.
According to state-backed newspaper The Paper, Seed was the first Chinese frontier AI laboratory to establish a standalone department dedicated to safety and alignment. The publication reported that the department is intended to improve AI-agent safety and address risks associated with advanced capabilities, including chemical, biological, radiological and nuclear (CBRN) risks.
The move comes amid wider discussions between Washington and Beijing about AI safeguards, emerging risks and channels for communicating about major technology-related incidents, even as the two countries compete for leadership in advanced AI.
ByteDance Launches Global Hiring Drive
In a recruitment announcement published on its official WeChat account on Wednesday, Seed said safety and alignment were becoming essential to determining whether increasingly capable AI models could be deployed at scale.
ByteDance steps up AI safety research
— WION (@WIONews) October 16, 2026
Autonomous AI systems raise concerns over unpredictable actions
Research to include automated testing and behaviour monitoring @AadhyaJunejaa brings you more pic.twitter.com/EfJNoWR5Ww
The research unit plans to recruit 10 researchers for positions in Beijing, Shanghai and San Jose, California. Seed has said the new team will receive computing capacity and research support comparable to those available to frontier-model development teams, allowing it to pursue safety research alongside the development of advanced AI capabilities.
New Team to Test AI Agents for Threats
The team will work across the AI development process, focusing on safety protocols, automated vulnerability testing, tools to monitor and explain model behaviour, and simulated environments for testing autonomous agents against potential threats. Its research will also cover safety evaluation, alignment training and methods for identifying and defending against emerging risks.
The work is intended to strengthen safeguards as AI capabilities advance, rather than respond to a specific rogue-AI incident involving ByteDance that has been established by the sources.
The focus on testing and monitoring reflects the challenge of keeping increasingly autonomous systems within appropriate safety limits.
Independent Audits to Strengthen AI Safeguards
Seed also plans to pursue independent third-party evaluations and work with universities and research institutions on AI safety. These are planned measures; the announcement does not establish that independent assessments have already verified the safety of its models.
The department's planned responsibilities include developing model specifications and safety policies, building evaluation frameworks, improving alignment training, strengthening protective systems and monitoring operational risks.
Seed says it intends to integrate safety research into model training, evaluation and deployment, while working with external partners to validate and improve model safety.
The Paper also reported that members of Seed's existing Responsible AI team and some Chat team staff would transfer to the new department. ByteDance launched its Seed research unit in 2023 to advance AI research across areas including multimodal models, media generation, scientific applications and robotics.
The new Alignment & Preparedness team adds a dedicated safety focus to that broader research programme.
