当前，人工智能正加速发展，且广泛出现于社会生活的各个方面。对此，各国政府愈加重视，积极出台政策和规范性文件以构建前瞻性布局。然而，人工智能在带来技术红利的同时，也在技术、伦理、道德、法律、政治等方面引发了风险，特别是法律层面，人工智能相关犯罪在我国已有实例。从研究现状看，目前研究常以某一具体应用场景为对象，区分其作为犯罪主体和犯罪工具来分析，论证不同情况下的责任归属问题，尚未形成立体化、系统化的视角和思路。而立法上，刑法的规制也并不完善，存在较多适用困境和立法空白。基于此，本文以网络安全为分析视阈，以系统安全、在线内容安全、组织管理安全这三层次为分析框架，将学术、立法、执法、产业等纳入同一体系中，从分析人工智能犯罪样态和特质，到探讨现行规则应对与应然规则建构，再到完善执法监管机构能力，和引导人工智能产业加强风险防控，使各方主体共同助力来确保人工智能安全、可靠、可控发展。 本文分四个章节，按照说明相关概念和研究方式、解读人工智能犯罪、研讨立法如何规制和实践中如何进行风险防控来展开。第一章，笔者简述了人工智能的概念、特征、技术应用现状及未来发展态势，以说明对其尽早加强治理的必要性。接着，通过阐述人工智能与网络安全的关系以确定本文分析视阈，通过说明人工智能不同分类以确定将弱人工智能、专用人工智能作为研究对象，通过比较确定以三安全分析框架为研究视角；第二章，笔者解释了人工智能犯罪的概念，指出从计算机犯罪到网络犯罪再到人工智能犯罪其实是一脉发展的结果，后按照三安全分析框架解读涉系统、数据、算法三种要素的侵害行为，并选取智能驾驶、模式识别、智能机器人这三个典型应用场景来对应阐述；第三章，现实需求、立法经验和政策体现的立法动向为刑事规范体系建构打下了基础，确定现阶段更适合采用一元的刑法典模式、渐进式与前瞻式相结合的立法形式后，在三安全分析框架下，一方面通过梳理和分析我国现有相关立法规范，用修订或增加司法解释的方式应对人工智能犯罪，另一方面通过构想和设计未来相关的立法规范，从国际治理和国内治理两个角度制定新的立法条款而加以规制；第四章，笔者建议执法领域促进机构组织的适应性变革、提升打击人工智能犯罪的专业能力和完善现有执法模式。同时建议产业领域在安全和发展两大价值导向下，增强对系统安全性的保护力，完善对在线内容的管理控制，提高对组织管理的专业认识，从而为应对人工智能犯罪开展全面的风险防控部署。

At present, Artificial Intelligence(AI) is developing rapidly, and widely appearing in all aspects of social life. In this regard, governments are paying more and more attention, and issuing relevant policies and normative documents actively to build a forward-looking layout.However, AI not only brings technology dividends, but also brings risks in Technology, Ethics, Morality, Law and Politics, especially in Law. There are some examples of Artificial Intelligence Crimes(AIC) in China.Looking from the current research situation, the researches on AI are often based on a specific application scene, distinguishing AI as a crime entity or a crime tool to analyze, demonstrating the attribution of responsibility in different situations, and have not yet formed a three-dimensional, systematic perspectives and ideas.In legislation, the regulation of this problem in Criminal Law is not perfect, there are more applicable difficulties and legislative gaps.Based on these, this paper takes Cyber Security as the analysis threshold, takes academic, legislative, law enforcement, industry into the Three-Security Analysis Framework (System Security, Online Content Security, Organization and Management Security). From the analysis of the patterns and characteristics of AIC to the discussion of the existing rules and should-be rules, and from the improvement of the ability of law enforcement to the guidance of AI industry in strengthening risk prevention and control, all parties need to be together to help ensure the safe, reliable and controllable development of AI. This paper is divided into four chapters, including the description of concepts and research methods, the interpretation of AIC, the discussion about how to regulate legislation and how to carry out risk prevention and control in practice.In the first chapter, the author briefly described the concept, characteristics, technical application status and future development trend of AI, in order to explain the necessity of strengthening governance as soon as possible. Then, the author determined the analysis threshold of this paper by expounding the relationship between AI and cyber security, determined Weak AI, Special AI as the research objects by explaining the different classifications of AI, and through the comparison, the Three-Security Analysis Framework was taken as the research perspective; In the second chapter, the author explained the concept of AIC, pointing out that from Computer Crime to Cyber Crime and to AIC is actually the result of a vein development. Then, the author interpreted the injurious acts about system, data and algorithm, according to the Three-Security Analysis Framework, and then chose three typical application scenarios of AI, the intelligent driving, the pattern recognition and the robot to describe in detail; In the third chapter, the author analyzed the actual demand, the legislative experience and the legislative tendency reflected by the policy, which laid the foundation for the construction of the criminal norms system. Then, the author determined that it is more suitable to adopt a monist model of Criminal Law, and a combination of progressive and forward-looking legislative forms. Under the Three-Security Analysis Framework, on the one hand, we can sort out and analyze the existing relevant legislative norms, to deal with AIC by revising or increasing judicial interpretation, on the other hand, we can conceive and design future relevant legislative norms, making new legislative provisions from international and domestic governance perspectives; In the fourth chapter, the author suggested that the law enforcement field should promote the adaptive change of the organization, enhance the professional ability of combating AIC and perfect the existing law enforcement mode. At the same time, it is suggested that under the guidance of the two major values of security and development, the industry should carry out a comprehensive risk prevention and control deployment, including strengthening the protection of system security, improving the management and control of online content, and enhancing the professional understanding of organizational management, in order to deal with AIC.