欧盟通过人工智能法案，加强大语言模型和AIGC监管
经过三年的协商，欧盟成员国终于批准了《人工智能法案》（AI Act）的最终文本，为人工智能领域的监管迈出了重要一步。
该法案旨在规范人工智能的使用，确保其安全、可信和尊重基本权利。它将人工智能系统分为四类，根据其风险水平进行监管。
高风险人工智能系统，例如用于医疗或交通的系统，将受到最严格的监管。它们必须经过严格的评估，并获得认证才能投入使用。
中风险人工智能系统，例如用于招聘或信用评分的系统，将受到较少的监管。它们必须符合某些透明度和公平性要求。
低风险人工智能系统，例如用于聊天机器人或游戏的人工智能，将受到最少的监管。
针对近期大语言模型和生成式人工智能（AIGC）的快速发展，欧盟立法者在2022年末增加了针对这些领域的监管条文。
这些条文要求大语言模型和AIGC系统提供有关其训练数据和算法的信息，并允许用户对生成的内容提出异议。
欧盟委员会副主席玛格丽特·维斯塔格表示：“人工智能法案将为人工智能在欧盟的发展提供明确的规则，确保其安全、可信和尊重我们的价值观。”
该法案预计将于2024年生效。它将成为全球第一个全面的人工智能监管框架，有望对人工智能行业的未来产生重大影响。
英语如下：
**Headline:** EU Finalizes AI Act, Bringing Regulation to Large Language Models
**Keywords:** AI Act, Regulation, Large Language Models
**Body:**
The European Union has adopted the final text of the Artificial Intelligence Act (AI Act), taking a significant step toward regulating the field of artificial intelligence.
The act aims to govern the use of AI, ensuring it is safe, trustworthy, and respectful of fundamental rights. It categorizes AI systems into four risk levels, with corresponding regulatory requirements.
High-risk AI systems, such as those used inhealthcare or transportation, will face the strictest regulations. They will need to undergo rigorous assessments and obtain certification before being deployed.
Medium-risk AI systems, such as those used in recruitment or credit scoring, will face less stringent regulations. They will need to comply with certain transparency and fairness requirements.
Low-risk AI systems, such as those used in chatbots or games, will face minimal regulation.
In response to the recent surge in large language models and generative AI (AIGC), EU lawmakers added provisions to the act in late 2022 to address these areas.
These provisions require large language models and AIGC systems to provide information about their training data and algorithms, and allow users to contest generated content.
“The AI Act will provide clear rules for the development of AI in the EU, ensuring it is safe, trustworthy, and respectful of our values,” said Margrethe Vestager, Executive Vice-President of the European Commission.
The act is expected to come into force in 2024. It will become the first comprehensive AI regulatory framework in the world and is likely to have a major impact on the future of the AI industry.
【来源】https://www.cls.cn/detail/1589217
Views: 2