China’s regulatory strategy in the global open-source AI landscape

May combine two claims.

As open-source model performance approached top closed-source models, US dependence on open-source models greatly increased, with a 62% usage rate, the highest globally.

“随着 开源模型性能向顶尖闭源模型靠拢 [33] ,美国对开 源模型的依赖度大幅提升,以 62% 的使用率位居 全球首位 [39] 。”

• cites [33]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [39]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

France's Hugging Face became the world's most important open-source AI platform, hosting more than 100,000 models in the first half of 2023.

“法国的 Hugging Face 已经成为全球最重要的开源 AI 平台,仅 2023 年上半年托管的模型数量就超过 10 万个 [51] 。”

• cites [51]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

Unlike Europe and the US, Chinese public opinion has consistently been highly positive toward AI, rising from 78% in 2022 to 83% in 2024.

“我国公众对 AI 技术始终保持着高度的积 极态度( 从 2022 年的 78% 到 2024 年的 83% )”

• cites [32]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

In the economic field, the position depends on the power balance: open-source dominance pushes toward incentive, closed-source dominance toward restriction, and parity toward the middle.

“该场域立场取决于两方的力 量对比:当开源方占优时,经济场域偏向激励端; 当闭源派占优时偏向限制方;当势均力敌时,趋向 中间。”

May combine two claims.

EU industry was constrained by backward industrial development, so the economic field's weight was relatively small, but it clearly pointed toward incentive.

“受限于落后的产业发展 [50] ,经济场域的权重占 比较小,但其内部明确指向激励端。”

• cites [50]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

In the social field, when external conditions are calm, no major security incident occurs, and transparency advantages are recognized, the game is at a middle state; if controllable risk mitigation measures exist, it tends toward incentive; during a major safety crisis, the pendulum rapidly shifts to restriction.

“社会场域 主体立场具有条件性与摇摆性:当外部条件平静, 无重大安全事件且透明度优势获得认可时,博弈 处于中间态;若此时已有可控风险缓解措施,则博 弈倾向激励方向;而在重大安全事故、负面案例曝 光的安全危机期,钟摆会急速偏向限制方。”

May combine two claims.

US regulatory practice has closely matched the Pendulum Model's predictions and reached equilibrium among pluralistic interests.

“The United States' regulatory practice aligns highly with the predictions of the Pendulum Model, thereby achieving equilibrium amid pluralistic interests.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Open-source AI is recognized as increasing model transparency and enabling beneficial regulation, thereby preventing misuse or attack.

“开源被公认为可增加模型透明度、向善 监管 [15] ,从而防止模型被滥用或受到攻击 [16] 。”

• cites [15]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [16]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

In the political field, when a country holds technological and industrial advantage, the game concerns standard dominance and technology spillover; when it is technologically weak and depends on foreign open-source models, the pendulum may shift toward restriction to prevent backdoor risks.

“当本国占据技术和产业优势时,涉及标准 主导权和技术外溢的博弈衡量;当本国处于技术 弱势端,依赖国外开源模型时,推动钟摆可能摆向 限制方向,以防范后门风险。”

May combine two claims.

The technical capability variable determines the weight of the social field; if open-source AI technology exceeds a threshold, its uncontrollable risk sharply amplifies the social field's weight, making public safety demands override other considerations.

“若开源 AI 的技术水平超乎一定阈值,其 不可控风险将急剧放大社会场域的权重,公众安 全诉求将压倒其他考量。”

May combine two claims.

China's open-source model catch-up intensified geopolitical tension and triggered a fierce internal political-field game in the US between standard dominance and technology leakage.

“中 国开源模型的追赶 [42] 奠定了地缘政治的紧张基 调,触发了场域内部的激烈博弈”

• cites [42]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

Chinese regulators remain wary of potential risks such as technology abuse and security hazards, so in the absence of obvious risk events, China's social-field game remains at the middle.

“但监管机构对潜在风险( 技术滥用、安全隐患) 保 持警惕。 因此,无论技术跃迁引发的权重如何变 化,在尚无明显风险事件时,我国社会场域的博弈 始终处于中间端。”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

Open-source AI is one of the key drivers for the rapid development of China's AI industry and is strategically important for achieving core technology autonomy and seizing future industrial high ground.

“开源 AI 无疑是推动我国 AI 产业快速发展的 关键动因之一,对于加快实现核心技术自主可控、 抢占未来产业制高点具有重要战略意义。”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

The risks accompanying open-source AI, including technology leakage, security and regulation, and backdoor intrusion, cannot be ignored.

“然而, 随之而来的技术泄露、安全和监管、后门侵入等风 险同样不可忽视。”

May combine two claims.

China's economic field has jumped to the dominant field with changes in industrial ecology and clearly swings toward incentive.

“经济场域 随 着 产 业 生 态 变 化 跃 升 为 主 导 场 域,并明确摆向激励端。”

May combine two claims.

In China's political field, although restriction forces from technology leakage and backdoor implantation exist, the field game still leans toward incentive because the technology gap with the US is small and backdoor risks have other solutions.

“在政治场域,尽管存在技术泄露和植入 后门的限制力量,但由于与美国的技术差距较小 (泄漏风险后果有限)且后门风险存在其他解决方 法,场域博弈结果仍偏向激励端。”

May combine two claims.

Synthesizing the field games, the Pendulum Model's deduced ought direction clearly swings to the incentive end, but contingency plans are needed for bidirectional security risks in the social and political fields.

“综合各场域的博弈力量,钟摆模型推导的应 然走向明确摆向激励端,但需同时为社会及政治 场域的双向安全风险提供应 对 备 案。”

May combine two claims.

In the EU, the economic field became dominant and highly incentive-oriented because local industry scale and open-source dependence greatly increased; Mistral AI's open-source models proved potential to compete directly with China and the US, and Hugging Face became a de facto global open-source AI resource center deeply bound to the EU.

“欧盟本土产业规模和开源依赖度 都大幅提升,产业生态变量发生较大变化:Mistral AI 的开源模型有力证明其具备与中美直接竞争的 潜力 [42] ;Hugging Face 以百万级开源规模而成为 全球事实上的开源 AI 资源中心 [52] ,与欧盟形成深 度绑 定 关 系。”

• cites [42]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [52]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.