À l’ONU, les patrons de l’IA prennent place à la table des États, mais qui écrira les règles ?
9月23日，联合国安理会首次就人工智能风险举行专门听证，OpenAI、Anthropic 和 Hugging Face 的负责人受邀发言，显示生成式 AI 已正式进入国际安全与治理议程。多国欧洲代表在会上主张，应由公共部门对这类技术实施更强监管，而不是完全交由企业自我约束。
这场讨论的一个显著特点是，受邀陈述分析意见的“专家”阵容中，没有任何一家欧洲本土、能够开发同类大模型的企业，也没有欧盟机构代表出席。换言之，围绕 AI 规则制定的核心对话，仍主要由美国公司及少数非欧洲参与者主导。
这一安排也折射出当前 AI 治理的现实矛盾：一方面，各国政府已将 AI 视为可能影响安全、信息环境和社会稳定的关键技术；另一方面，真正掌握模型能力、算力与产品落地经验的，仍是少数大型科技公司。欧洲国家虽然强调公共监管，但在技术与产业层面缺乏同等重量级的代表，意味着未来规则的制定权、解释权和执行标准，可能继续向美国企业与国际机构倾斜。
On 23 September, the UN Security Council heard from the leaders of OpenAI, Anthropic and Hugging Face on the risks posed by artificial intelligence. The session marked another sign that AI companies are now being brought directly into discussions once reserved for states, as governments seek to understand and shape the technology’s impact.
Several European countries used the occasion to argue for public oversight of AI systems, insisting that these technologies should not be left to private actors alone. But the lineup of invited experts also highlighted a notable imbalance: no European company developing comparable foundation models was present, and no institution from the European Union was included among the analysts asked to brief the Council.
The debate underscored a broader question now facing international institutions: if AI firms are increasingly treated as key interlocutors by governments, who will actually write the rules governing them?
Hier, 23 septembre, le Conseil de sécurité a entendu les dirigeants d’OpenAI, d’Anthropic et de Hugging Face sur les risques de l’intelligence artificielle. Plusieurs États européens ont défendu un contrôle public de ces technologies. Mais parmi les experts invités à présenter leur analyse, aucune entreprise européenne développant des modèles comparables, ni aucune institution de l’Union …
L’article À l’ONU, les patrons de l’IA prennent place à la table des États, mais qui écrira les règles ? est apparu en premier sur FRENCHWEB.FR : innovation, tech, network.
The UN Security Council heard OpenAI, Anthropic, and Hugging Face leaders on AI risks, but European public-control advocates noted a lack of comparable European model builders and EU institutions among invited experts—raising questions about who will shape AI rules.
OpenAI — Builds frontier AI models; US-based.
Anthropic — Develops AI safety-focused models; US-based.
Hugging Face — Provides open AI model tooling and hosting; France-based.
[Monitor] — UN and European stakeholders’ push for public oversight could reshape regulation and market access for AI model developers.