贵州茅台酒股份有限公司负责任人工智能政策

贵州茅台酒股份有限公司负责任人工智能政策

  本政策旨在规范贵州茅台酒股份有限公司（以下简称“本公司”）在人工智能技术研发、部署与应用过程中的行为准则，确保人工智能技术的应用符合法律法规、社会伦理与行业最佳实践。本公司深刻认识到人工智能技术为酒类行业带来的创新机遇与潜在风险，承诺以负责任的态度推动人工智能在营销推广、客户服务、供应链管理、产品研发等领域的应用，在提升效率与用户体验的同时，切实保护利益相关方的合法权益。

  本公司将遵循本公司“共生 共享 共赢”的ESG理念，将负责任人工智能融入企业治理体系，并依据《中华人民共和国网络安全法》《中华人民共和国数据安全法》《中华人民共和国个人信息保护法》《生成式人工智能服务管理暂行办法》等法律法规规章，制定本政策。

  一、适用范围

  本政策适用于本公司在业务运营中采购、部署或使用人工智能技术的所有场景，包括但不限于：

  1.面向消费者的AI驱动营销推广与个性化推荐系统；

  2.客户服务中的智能对话机器人与虚拟助手；

  3.公司内部管理中的AI辅助决策工具；

  4.利用生成式人工智能技术生成文本、图片、音视频等内容。

  为本公司提供AI产品、服务或技术的第三方供应商与合作伙伴，在涉及本公司业务时应参照本政策执行。

  二、人工智能治理基本原则

  本公司在人工智能技术的全生命周期中，遵循以下六项基本原则：

  （一）合法合规

  所有AI应用须严格遵守适用的法律法规规章与行业规范，包括但不限于网络安全、数据保护、消费者权益保护以及酒类广告、营销等方面的特殊规定。

  （二）安全可控

  AI系统应具备稳健的安全防护能力，防范恶意攻击与滥用风险，确保系统在预期范围内稳定运行，且始终保持在人类有效控制之下。

  （三）公平无歧视

  本地部署的大模型在微调数据选择、决策输出等环节，采取有效措施防止基于民族、信仰、健康状况等因素的歧视，促进结果公平。

  （四）透明可解释

  AI应用的功能应被清晰告知，涉及关键决策时应提供可理解的解释。使用生成式AI技术产生的内容须按照相关法律法规进行显式标识或隐式标识。

  （五）可追溯与可问责

  AI应用的研发、部署与运行过程应建立完整的记录与审计机制，明确各环节的责任主体，确保AI行为及其影响可追溯、可问责。

  （六）数据最小化与隐私保护

  AI应用对个人信息的收集与处理应遵循合法、正当、必要原则，采取加密、去标识化、匿名化等技术手段保障数据安全，尊重用户对其个人信息的控制权。

  三、数据隐私与安全保护

  （一）个人信息保护

  AI系统在收集、使用、存储、传输个人信息时，应当按照法律法规的规定确定处理的合法性基础。除法律法规规定不需取得个人同意的情形外，应当取得信息主体的明确同意，并以清晰易懂的方式告知信息收集的目的、方式与范围。对个人权益有重大影响的个人信息处理活动时，应当按照法律法规的规定，事前进行个人信息保护影响评估，并对处理情况进行记录。

  （二）网络安全保障

  AI系统应建立多层安全防护体系，包括但不限于：输入数据安全校验、输出内容安全过滤等。

  （三）训练数据合规

  当开展AI模型的微调训练时，训练所使用的数据应具有合法来源。涉及知识产权的训练数据，须确保不侵犯他人合法权益。涉及个人信息的，须依法取得个人同意。应采取措施提高训练数据的真实性、准确性、客观性与多样性。

  四、算法公平与反歧视

  AI系统在训练数据选择、模型优化与提供服务的过程中，须采取有效措施防止各类歧视。包括但不限于以下方式：

  1.确保训练数据的代表性，避免因数据偏差导致族群、地域、性别等方面的不公；

  2.在酒类营销AI系统中，确保算法不诱导消费者过量饮酒；

  3.建立算法偏见的发现与纠正机制，接受内外部监督。

  具体而言，不得利用算法放大酒类消费的社交压力，不得以虚构或夸大的方式宣传酒类产品的功效、年份、产地或奖项等。

  五、透明度与可解释性

  （一）AI交互告知

  当用户与AI系统（如智能客服、内容生成工具）进行交互时，应以适当方式告知用户其正在与AI系统交互，而非人类。

  （二）能力边界说明

  AI服务的适用人群、场合、用途以及技术局限性应当被明确告知用户，避免用户对AI能力产生不合理的期望或依赖。

  六、人类监督与问责机制

  （一）人工干预机制

  涉及关键决策场景的AI应用，AI仅提供辅助建议，最终决策权保留给具备相应权限的人类。AI系统应设计为在必要时人类可介入、修正或终止AI的决策或行为。

  （二）可追溯记录

  AI系统的关键决策过程、基础模型来源、模型变更记录、人工干预记录等均须完整保存，保存期限不得低于法律法规要求。相关记录应可供内外部审计查阅。

  七、AI使用边界与禁止事项

  本公司在AI应用中严格界定使用边界，明确以下禁止行为：

  1.不得使用AI技术实施操纵消费者行为、利用消费者认知弱点进行不当营销；

  2.不得部署涉及社会评分、大规模未经授权的生物特征监控的AI系统；

  3.不得使用AI生成或传播危害国家安全、破坏社会稳定、宣扬恐怖主义、暴力、淫秽色情及虚假有害信息；

  4.不得利用AI算法实施垄断、不正当竞争行为或侵犯他人知识产权；

  5.不得将AI决策用于可能对个人重大财产权益产生直接风险的场景；

  6.不得使用AI技术规避法律法规对酒类广告与营销的限制性要求。

  本公司还将建立AI风险评估与准入机制，所有AI项目上线前须完成合规性评估，高风险AI应用须经审批后方可部署。涉及面向公众提供生成式AI服务的，须按国家有关规定完成安全评估与互联网信息服务算法备案手续。

  八、环境可持续

  本公司关注人工智能技术的环境足迹，承诺在AI基础设施建设与模型训练过程中践行绿色低碳理念。在自有或第三方AI数据中心的选择与使用中，优先采用高能效计算设备与可再生能源，优化模型架构以降低算力消耗。

  九、未成年人保护

  根据《中华人民共和国未成年人保护法》及酒类行业监管要求，本公司严格禁止向未满十八周岁的未成年人销售酒类产品。AI系统在涉及酒类产品展示、推广、销售等环节，均须嵌入年龄验证与过滤机制，不向未成年用户推送AI生成的酒类相关内容。

  十、政策更新

  本公司保留根据需要修订本政策的权利。当法律法规规章、行业规范等发生重大变化，或公司AI应用范围发生实质性扩展时，将及时对本政策进行评估与更新。重要更新将通过公司官方官网等方式予以公告。

  定义

  除非上下文另有说明，本政策中使用的术语定义如下：

  人工智能（AI）：指使机器能够模拟人类智能的技术，包括但不限于机器学习、深度学习、自然语言处理、计算机视觉、生成式人工智能等。

  生成式人工智能：指具有文本、图片、音频、视频等内容生成能力的模型及相关技术。

  AI系统：指基于人工智能技术构建的软件系统、平台或工具。

  个人信息：指以电子或者其他方式记录的与已识别或者可识别自然人有关的各种信息，不包括匿名化处理后的信息。

  Kweichow Moutai Co., Ltd.

  Responsible Artificial Intelligence Policy

  This Policy is intended to regulate the code of conduct of Kweichow Moutai Co., Ltd. (hereinafter referred to as "the Company") in the research and development, deployment and application of artificial intelligence technology, so as to ensure that the application of AI technology complies with laws and regulations, social ethics and industry best practices. The Company is fully aware of the innovation opportunities and potential risks that AI technology brings to the alcoholic beverage industry, and is committed to promoting the application of AI in areas such as marketing and promotion, customer service, supply chain management and product research and development in a responsible manner, while enhancing efficiency and user experience and effectively protecting the legitimate rights and interests of stakeholders.

  The Company will follow the ESG philosophy of "Symbiotic Coexistence, Sharing Benefits, Mutual Success" to integrate responsible AI into its corporate governance system, and formulate this Policy in accordance with the Cybersecurity Law of the People"s Republic of China, the Data Security Law of the People"s Republic of China, the Personal Information Protection Law of the People"s Republic of China, the Interim Measures for the Administration of Generative Artificial Intelligence Services and other laws, regulations and rules.

  I. Scope of Application

  This Policy applies to all scenarios in which the Company procures, deploys or uses AI technology in its business operations, including but not limited to:

  1. Consumer-facing AI-driven marketing promotion and personalized recommendation system;

  2. Intelligent chatbots and virtual assistants in customer service;

  3. AI-assisted decision-making tools in internal corporate management;

  4. The use of generative artificial intelligence technology to generate content such as text, images, audio and video.

  Third-party suppliers and partners that provide AI products, services or technologies to the Company shall, to the extent that their business involves the Company, apply this Policy mutatis mutandis.

  II. Basic Principles of AI Governance

  In the full life cycle of AI technology, the Company follows the following six basic principles:

  (1) Lawful Compliance

  All AI applications must strictly comply with applicable laws, regulations, rules and industry norms, including but not limited to the special requirements concerning cybersecurity, data protection, consumer rights protection, and alcoholic beverage advertising and marketing.

  (2) Safety and Controllability

  AI systems shall be equipped with robust security protection capabilities to prevent risks of malicious attacks and misuse, ensure that the systems operate stably within the expected scope, and always remain under effective human control.

  (3) Fairness and Non-Discrimination

  In such aspects as the selection of fine-tuning data and the output of decisions, locally deployed large models shall take effective measures to prevent discrimination based on factors such as ethnicity, belief and health status, and to promote fair outcomes.

  (4) Transparency and Explainability

  The functions of AI applications shall be clearly disclosed, and comprehensible explanations shall be provided when critical decisions are involved. Content generated using generative AI technology shall be explicitly or implicitly labeled in accordance with relevant laws and regulations.

  (5) Traceability and Accountability

  Complete records and audit mechanisms shall be established for the research and development, deployment and operation of AI applications, and the parties responsible for each stage shall be clearly identified, so as to ensure that the behaviors of AI and their impacts are traceable and accountable.

  (6) Data Minimization and Privacy Protection

  The collection and processing of personal information by AI applications shall follow the principles of lawfulness, legitimacy and necessity, adopt technical measures such as encryption, de-identification and anonymization to safeguard data security, and respect users" control over their own personal information.

  III. Data Privacy and Security Protection

  (1) Protection of Personal Information

  When collecting, using, storing, or transmitting personal information, AI systems shall determine the legal basis for processing in accordance with laws and regulations. Except in circumstances where laws and regulations provide that individual consent is not required, the explicit consent of the information subject shall be obtained, and the purpose, method, and scope of information collection shall be communicated in a clear and understandable manner. For personal information processing activities that have a significant impact on individual rights and interests, a personal information protection impact assessment shall be conducted in advance in accordance with laws and regulations, and the processing activities shall be recorded.

  (2) Cybersecurity Protection

  AI systems shall establish a multi-layered security protection system, including but not limited to: input data security validation, output content security filtering, etc.

  (3) Compliance of Training Data

  When fine-tuning training is carried out on AI models, the data used for training shall have lawful sources. The Company shall ensure that training data involving intellectual property rights does not infringe upon the legitimate rights and interests of others. Training data involving personal information shall be subject to the consent of the individuals obtained in accordance with the law. Measures shall be taken to improve the authenticity, accuracy, objectivity and diversity of training data.

  IV. Algorithmic Fairness and Anti-Discrimination

  In the process of training data selection, model optimization and service provision, AI systems shall take effective measures to prevent various types of discrimination, including but not limited to the following ways:

  1. Ensure the representativeness of training data to avoid unfairness with respect to ethnic groups, regions, gender and other aspects caused by data bias;

  2. In alcoholic beverage marketing AI systems, ensure that algorithms do not induce consumers to drink excessively;

  3. Establish mechanisms for the detection and correction of algorithmic bias, and accept internal and external supervision.

  Specifically, algorithms shall not be used to amplify the social pressure of alcoholic beverage consumption, and the efficacy, vintage, origin or awards of alcoholic beverage products shall not be publicized in a fabricated or exaggerated manner.

  V. Transparency and Explainability

  (1) Disclosure of AI Interaction

  When users interact with AI systems (such as intelligent customer service and content generation tools), they shall be informed in an appropriate manner that they are interacting with an AI system rather than a human.

  (2) Explanation of Capability Boundaries

  The applicable population, occasions, purposes and technical limitations of AI services shall be clearly communicated to users, so as to prevent users from forming unreasonable expectations of or dependence on AI capabilities.

  VI. Human Oversight and Accountability Mechanisms

  (1) Human Intervention Mechanism

  For AI applications involving key decision-making scenarios, AI shall only provide auxiliary suggestions, and the final decision-making authority shall be reserved for humans with the appropriate authorization. AI systems shall be designed so that humans can intervene in, correct or terminate the decisions or behaviors of AI when necessary.

  (2) Traceable Records

  The key decision-making processes of AI systems, the sources of base models, records of model changes and records of human intervention shall all be completely preserved, and the retention period shall not be shorter than that required by laws and regulations. The relevant records shall be available for review in internal and external audits.

  VII. Boundaries of AI Use and Prohibited Matters

  The Company strictly defines the boundaries of AI use in its AI applications and explicitly sets out the following prohibited acts:

  1. AI technology shall not be used to manipulate consumer behavior or to engage in improper marketing by exploiting consumers" cognitive weaknesses;

  2. AI systems involving social scoring or large-scale unauthorized biometric surveillance shall not be deployed;

  3. AI shall not be used to generate or disseminate information that endangers national security, undermines social stability, advocates terrorism or violence, or contains obscene or pornographic content, or false and harmful information;

  4. AI algorithms shall not be used to implement monopolistic or unfair competition practices, or to infringe upon the intellectual property rights of others;

  5. AI decisions shall not be used in scenarios that may pose direct risks to the significant property rights and interests of individuals;

  6. AI technology shall not be used to circumvent the restrictive requirements of laws and regulations on alcoholic beverage advertising and marketing.

  The Company will also establish an AI risk assessment and admission mechanism. All AI projects shall complete a compliance assessment before launch, and high-risk AI applications shall be deployed only after approval. Where generative AI services are provided to the public, the security assessment and the internet information service algorithm filing procedures shall be completed in accordance with relevant national regulations.

  VIII. Environmental Sustainability

  The Company pays attention to the environmental footprint of AI technology and is committed to practicing the concept of green and low-carbon development in the construction of AI infrastructure and the process of model training. In the selection and use of self-operated or third-party AI data centers, the Company will give priority to the adoption of high-energy-efficiency computing equipment and renewable energy, and optimize model architectures to reduce computing power consumption.

  IX. Protection of Minors

  In accordance with the Law of the People"s Republic of China on the Protection of Minors and the regulatory requirements of the alcoholic beverage industry, the Company strictly prohibits the sale of alcoholic beverage products to minors under the age of eighteen. In all aspects involving the display, promotion and sale of alcoholic beverage products, AI systems shall incorporate age verification and filtering mechanisms and shall not push AI-generated alcoholic beverage-related content to underage users.

  X. Policy Updates

  The Company reserves the right to revise this Policy as needed. When major changes occur in laws, regulations, rules and industry norms, or when the scope of the Company"s AI applications is substantially expanded, this Policy shall be assessed and updated in a timely manner. Important updates shall be announced through the Company"s official website and other channels.

  Definitions

  Unless the context otherwise indicates, the terms used in this Policy are defined as follows:

  Artificial Intelligence (AI): refers to technologies that enable machines to simulate human intelligence, including but not limited to machine learning, deep learning, natural language processing, computer vision, and generative artificial intelligence.

  Generative Artificial Intelligence: refers to models and related technologies capable of generating content such as text, images, audio and video.

  AI System: refers to software systems, platforms or tools built on artificial intelligence technology.

  Personal Information: refers to all kinds of information recorded electronically or in other ways related to identified or identifiable natural persons, excluding information that has been anonymized.