《欧盟人工智能法案》的背景、主要内容与评价——兼论该法案对劳动法的影响

Background,Main Contents,and Evaluation of the EU Artificial Intelligence Act–Also on Its Implications for Labor Law

为了应对人工智能系统带来的各种挑战,《欧盟人工智能法案》采用基于风险分类的监管路径。法案首先明确禁止了某些存在不可接受风险的人工智能系统,然后重点规制高风险人工智能系统,任何在欧盟市场上提供此类系统的企业都必须满足法案规定的从风险管理、数据治理到全面的记录义务等诸多要求。但是,法案的监管要求可能造成小微企业负担过重,模糊的规定也可能导致实施方面的困难。在劳动关系领域,对雇员进行社会评分、预测雇员犯罪风险、分析雇员情绪等人工智能系统可能因为存在不可接受的风险而被禁止。对岗位申请者和雇员使用人工智能系统的,则根据是否属于高风险系统而负担不同的义务。除了要让雇主向雇员提供相关人工智能技能培训以外,工作中使用ChatGPT的问题也值得关注。

The Parliament of the EU adopted the Artificial Intelligence Act in March 2024,which takes a risk-based approach.The Act establishes obligations for providers and users depending on the level of risk from artificial intelligence.It prohibits certain AI systems associated with an unacceptable risk.AI systems that negatively affect safety or fundamental rights will be considered high risk.All high-risk AI systems will be assessed before being put on the market and throughout their lifecycle.Anyone offering such systems on the market must comply with numerous requirements,ranging from risk management and data governance to comprehensive documentation obligations.The“operator”of a high-risk AI system is first and foremost obliged to comply with the mandatory instructions for use provided by the provider.Special rules exist for systems with a“wide scope of application”such as ChatGPT.When using systems that do not entail high risk,only a few provisions that apply to all AI systems need to be observed.In the event of violations,high penalties of up to 7%of global turnover are envisaged.The AI Act does not address the question of which training material the developer of AI can use without violating data protection or copyright law.In this respect,the GDPR and the relevant copyright regulations apply.The handling of machine-related data has also not been regulated.These problems could significantly hinder the further development of AI.Additionally,there is no consideration of what will happen if entire professions such as journalists or translators are reduced to just a few people.The current labor laws of the EU member states also do not have any effective means of cushioning a slump in employment in the event of widespread use of AI.A responsible legislator should have faced up to the problems in the labor market.The regulatory requirements of the AI Act may create an undue burden on MSMEs.Vague statements in the Regulation may also lead to difficulties in implementation.In the field of labor relations,the application of AI systems such as social scoring of employees,predicting the risk of employees committing crimes,and analyzing employees’emotions may be prohibited due to unacceptable risks.When AI systems are used in the recruitment procedure or human resource management,employers must fulfill different obligations,depending on whether the AI systems are considered high-risk systems.Since employers are obliged to ensure that employees have adequate AI skills,they must offer free training at their own expense.Issues related to the use of ChatGPT in work deserve further study.Unauthorized use of ChatGPT by employees would constitute a violation of labor contract obligations.It seems unwise,at least for the time being,to use ChatGPT in legal advice work,as the system is prone to error.