DeepSeek Reaffirms Open-Source Strategy for AGI
DeepSeek research engineer Gao Xi reportedly told the 2026 Wuzhen Internet Conference that the company will continue promoting open source to accelerate the arrival of AGI rather than merely compete with other model developers. He said the team favors promptly disclosing more efficient training methods, model architectures, and inference techniques, while acknowledging that released models cannot be easily withdrawn. The report also says DeepSeek is strengthening security protections and considering model tiering as open-source governance remains incomplete.
DeepSeek is pairing its open-source position with attention to security risks and model governance, while the core claims about disclosure and AGI goals come from company representatives as reported by a single outlet.


DeepSeek 研究工程師高熙近日在 2026 年世界互聯網大會烏鎮峰會的 AI 論壇上重申,公司將堅定推動「開源」策略。他強調,此舉的核心目標是為了加速通用人工智慧(AGI)的到來,而非單純在模型競賽中與其他企業爭奪名次。高熙直言,隨著 AI 發展步伐持續加快,業界真正的競爭對手已非同業,而是「時間」。
高熙進一步說明,未來若在研發過程中發掘更高效的訓練方法、模型架構或推論技術,DeepSeek 團隊傾向在第一時間對外公開。他認為,開源的最大價值在於將最新的研究成果迅速交予廣大開發者進行測試與驗證,藉由匯聚社群力量共同促成技術突破,進而大幅縮短整個產業邁向 AGI 的時程。