ByteDance’s Seed Creates AI Safety Team, Seeks 10 Researchers

What changed

ByteDance’s AI research unit Seed has created an Alignment & Preparedness team and opened a global search for 10 researchers. The recruitment post, published on October 14, 2026, lists positions in Beijing, Shanghai and San Jose, with work spanning safety training, vulnerability testing, model monitoring, AI-agent testing and preparation for emerging risks.

Why it matters

Seed says the new team will receive computing capacity and research resources comparable to those given to frontier-model teams. That matters because safety work is being placed alongside model development, rather than treated as a final inspection before release.

The timing is significant. AI agents are gaining more autonomy, including the ability to access the internet, while scrutiny of their safety grows in both China and the United States. Seed’s plan covers the whole development cycle: training, evaluation and deployment. It also includes independent testing and collaboration with universities and research institutions.

My read is that the announcement makes safety expertise part of the competition for more capable models. The harder test will be whether the team gains enough authority, staff and resources to influence systems before they are deployed, rather than simply document problems afterward.