欧盟AI法案的核心规定从今年8月开始执行[1]，这标志着欧洲委员会正式成为全球最主要的人工智能监管机构[1]。该法案对大型语言模型等AI模型制定了一系列强制性要求，包括开发者必须披露训练数据中的受版权保护内容[1]，并向下游用户提供充分信息以便其了解模型的能力[1]。针对开发最强大"前沿模型"的企业，法案还要求其识别并缓解可能对社会造成的风险[1]。
这些监管规则不仅对欧盟内部的AI开发者适用，也适用于所有在欧盟市场商业化AI技术的外国公司[1]。欧洲委员会已建立欧洲AI办公室来推动这些规则的执行[1]，尽管该机构面临资源有限和AI人才竞争激烈等挑战[1]。欧盟这一监管模式预计将为全球其他地区树立标准，成为国际AI治理的基准[1]。
The European Union's landmark artificial intelligence law enters into force this August, positioning the European Commission as the world's primary AI regulator.[1] The regulations, passed in 2024, introduce comprehensive requirements for developers of large language models and other advanced AI systems operating within the EU market.[1]
Under the new framework, AI model developers must disclose copyrighted content used in training data and provide sufficient information enabling downstream users to understand model capabilities.[1] Companies developing the most powerful "frontier models" face particular obligations to identify and mitigate risks these systems may pose to society.[1] Importantly, the rules apply not only to EU-based companies but also to foreign organizations commercializing AI technology within the European market.[1]
The European Commission has established the European AI Office to oversee enforcement of these regulations, though it confronts significant obstacles including limited resources and intense competition for AI talent.[1] The EU's regulatory approach is expected to serve as a global benchmark, potentially shaping how other regions approach AI governance.[1]