不可变版本 / Immutable version 2026-10-25.3. 本副本不加载分析或应用脚本 / No analytics or app scripts.
适用于文本、图片、音频、视频及虚拟场景的生成、复制、下载和对外发布。
2026-10-25.3 · 2026-10-25
AI 标识用于说明内容由人工智能生成或合成，不表示平台证明内容真实、合法、安全或取得全部第三方授权。界面中的「AI」「AI 生成」等文字、符号或场景提示为显式提示；文件元数据中的生成属性、提供者和内容编号等为隐式标识。具体形式随输出格式、功能和适用规范而定，不能因未肉眼看到水印就认定内容不是 AI 生成。
Laintas 应按适用规范在相关界面及需标识的输出中提供提示，支持文件元数据的格式按要求处理隐式标识。复制、转码、压缩、裁切或导出可能改变标识，应保留或补充依法必要的说明。不同格式和来源具有不同标识能力，本文不是所有现有文件均已完成全部技术验证的保证；发现缺失可附文件来源和格式联系修正，不能把缺失当作允许冒充人工创作。
对外传播时，应如实说明依法需要披露的 AI 生成属性，并使用传播平台的标识功能。不得恶意删除、篡改、伪造、隐匿法定标识，或通过不当标识损害他人权益。Laintas 提示不替代您对发布内容的权利、事实、隐私、肖像、声音和行业限制核查，也不免除平台自身义务。
如某项服务依法允许提供无显式标识版本，须通过专门流程明确您的标识义务和使用责任，并按规定保留必要提供对象等日志；不得以提示词要求或普通导出操作绕过。本文不表示所有产品开放去除标识功能，也不授权去除依法仍需保留的隐式标识。对适用的日志保存义务，隐私政策的一般删除规则依法让位。
不得利用生成内容从事欺诈、未经许可的身份或声像仿冒、恶意诽谤及其他侵权。投诉可通过 [email protected] 或主站联系页提交内容链接、生成时间或任务号、权利依据与问题说明，不必公开敏感证据。我们按证据、风险和法律采取必要处置、核实及申诉处理。
人工润色、合并代码或二次编辑不必然消除依法需要披露的生成属性。纯人工资料、引用的第三方资料与模型新增部分应按实际情况说明，不能用 AI 标签冒领他人作品，也不能把有第三方权利的内容当作自由素材。Issues 中人工回答若实质使用模型生成内容，应如实披露，并遵守该问题是否允许 AI 作答的规则；仅把 Solver 输出转贴为“人工”不改变其来源。对外声明须与所发布部分和实际创作过程相称。
发现文件或页面标识遗漏、错误归因或被他人恶意改动时，可提供原文件、任务号、下载格式、发布地址及必要截图，避免再次公开敏感原始内容。平台应核对生成来源、复制或导出链路和适用格式要求，采取补充说明、纠正归因、修复导出或限制传播等相应措施。技术能力与格式差异不免除依法应完成的标识；未完成核查的文件不应被表述为已经获得合规认证。
For generating, copying, downloading and publishing text, images, audio, video and virtual scenes.
2026-10-25.3 · 2026-10-25
AI labels identify generated or synthesised content, not verified truth, legality, safety or third-party rights. “AI”, “AI-generated”, symbols and contextual notices are explicit indications; generation attributes, provider identifiers and content numbers in file metadata are implicit labels. Form depends on format, function and applicable rules. Absence of a visible watermark does not establish human origin.
Laintas must provide notices in relevant interfaces and outputs under applicable rules and address required implicit labels for metadata-capable formats. Copying, transcoding, compression, cropping or export can alter labels; preserve or restore required notices. Formats and sources differ, and this document is not certification that every existing file has passed every technical check. Report omissions with source and format for correction; an omission does not authorise pretending content is human-created.
Disclose AI origin where required and use the receiving platform’s labelling controls. Do not maliciously remove, alter, forge or hide required labels or misuse labels to harm rights. A Laintas notice does not replace checks of rights, facts, privacy, likeness, voice and industry restrictions, nor remove the platform’s own duties.
Where lawful unlabelled versions are available, a dedicated process must clarify your labelling duties and use responsibility and retain legally required recipient logs. Prompts or ordinary exports cannot bypass it. This does not offer label removal in every product or permit removal of implicit labels that remain required. Applicable log duties take priority over general deletion subject to law.
Do not use generated content for fraud, unauthorised identity, likeness or voice impersonation, malicious defamation or other infringement. Report a link, time or task ID, rights basis and explanation to [email protected] or Contact, without publishing sensitive evidence. Evidence, risk and law determine necessary action, verification and review.
Human polishing, code merging and editing do not necessarily remove required AI-origin disclosure. Describe human material, third-party quotations and model additions accurately. An AI label cannot claim another author’s work or turn protected material into free assets. Human Issues answers materially using generated content should disclose that use and follow the question’s AI restrictions. Relabelling Solver output as human does not change its origin. Statements should match the published material and actual creation process.
For omitted labels, false attribution or malicious alteration, provide the original file, task ID, format, publication link and necessary screenshots without republishing sensitive material. Laintas should check generation provenance, copying/export paths and format rules, and use appropriate explanations, attribution corrections, export fixes or distribution restrictions. Technical and format differences do not waive required labelling. Unverified files must not be described as compliance-certified.