Zhang Yiming, ByteDance founder, issued a rare public statement on AI strategy, escalating global competition. On August 5, The Information reported that at Seed’s all-hands meeting last month, Zhang set a strict rule for the large model R&D team: they must not rely on distilling other companies' cutting-edge models. Since the second half of 2024, Zhang has participated in Seed’s core technical team reviews monthly, now devoting roughly half his working time to Seed. In July 2026, during a team discussion on catching up via distillation, Zhang spoke up, stating that even if Seed falls behind temporarily, it cannot use distillation. Distillation, a technique systematically promoted by Geoffrey Hinton in 2015, involves using a stronger model as a teacher, but now raises intellectual property and Sino-US AI competition concerns. Inside Seed, the debate was straightforward: why not learn from others' answers? ByteDance faces pressure in large language models, especially when stronger open-source models appear. Zhang made the final call, saying to achieve long-term goals, they should sacrifice short-term interests. The Information partly attributed this decision to TikTok’s risks in the United States. On August 6, ByteDance CEO Liang Rubo responded at ByteDance’s mid-year all-hands meeting, admitting the statement.