Anthropic 2026 版使用政策禁止对 Claude 的极端且无谓残忍对待，并将选举、武器与监控条款成文化——11 月 12 日生效，沮丧、反驳、黑暗创作主题与模型测试被明确排除在外

“Anthropic 新使用政策 11 月 12 日生效——只有极端且无可辨识目的的残忍对待 Claude 会构成违规，沮丧、反驳、黑暗创作主题与模型测试明确不适用。”

可用说法

• 2026 年 10 月 8 日（星期四，已按日历核算），Anthropic 发布 2026 版使用政策更新，其中写明：「The updated policy takes effect on November 12.（更新后的政策于 11 月 12 日生效。）」按博文：「Most of the updates in the latest version are intended to clarify existing rules.（最新这一版的多数更新都是为了厘清既有规则。）」关于模型滥用，按博文：「We've added a prohibition on sustained and needless abusive or cruel behavior toward our models. The policy update is meant to apply only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose. It does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research.（我们新增了一条禁令：不得对模型实施持续且无必要的辱骂或残忍行为。该政策更新只打算适用于极端情形——用户反复对模型施加残忍行为且无任何可辨识目的。它不适用于常见的用户沮丧、反驳、黑暗创作主题，或模型测试与研究。）」按博文，执行方式保持对话内：「Such abuse is the main focus of this update; Claude's ability to end these interactions will remain the primary enforcement mechanism.（这种滥用是本次更新的主要焦点；Claude 结束这类对话的能力将仍是主要执行机制。）」，其基础是已落地的一步——「allowing Claude models to end rare conversations with persistently abusive users on Claude.ai and Claude Code（允许 Claude 模型在 Claude.ai 与 Claude Code 上结束与持续辱骂用户的少数对话）」。关于选举，按博文，该节被收窄聚焦到「disallowing the use of Claude to deceive voters or disrupt elections（禁止用 Claude 欺骗选民或破坏选举）」并更名为「Do Not Undermine Democratic Processes（不得破坏民主进程）」，且博文写明：「For a similar reason, we've also removed our blanket prohibition on personalized vote and campaign targeting. This prohibition covered legitimate civic work, such as nonprofits using Claude to write information for voters in other languages, or election officials sending ballot cure notices.（出于类似的原因，我们还移除了对个性化投票与竞选定向的一揽子禁令。该禁令此前覆盖合法的公民工作，比如非营利组织用 Claude 为选民撰写其他语言的信息，或选举官员发送选票补正通知。）」关于欺骗性活动，按博文：「We've consolidated those rules into a new section titled Do Not Engage in Deceptive Campaigns or Artificial Activity, which applies to deceptive activity of any kind (whether political or commercial). It covers efforts to obscure who is behind a message or amplify content through fake accounts or posts, along with building the tools and infrastructure for running influence operation campaigns.（我们把那些规则整合进一个新章节，题为 Do Not Engage in Deceptive Campaigns or Artificial Activity，适用于任何类型的欺骗性活动（无论政治还是商业）。它涵盖掩盖讯息背后主体、通过假账号或假帖文放大内容的行为，也包括搭建影响行动活动的工具与基础设施。）」关于武器，按博文：「The updated section makes clear that our prohibitions include the software and components that make weapons work, as well as actions like arming drones and other autonomous vehicles.（更新后的章节明确，我们的禁令包括让武器运转的软件与部件，以及为无人机等自主载具武装之类的行为。）」关于监控，按博文：「tracking people without their consent is prohibited, whether it happens in real time or through analysis of previously collected data. Claude cannot be used to decide or recommend who to investigate, arrest, or charge in a law enforcement or criminal justice process（未经同意追踪人属被禁止行为，无论是实时发生还是通过分析既有收集数据发生。Claude 不得被用于在执法或刑事司法程序中决定或建议调查、逮捕或起诉谁）」，而允许的用途包括「tracking that people have agreed to (such as fraud monitoring), content moderation, journalism, and legal research（人们已同意的追踪（如反欺诈监测）、内容审核、新闻工作与法律研究）」。对武器与监控两节，博文自己把改动定性为厘清：「this update does not reflect a change in what we enforce in practice: it is just intended to make our policy more explicit（本次更新并不反映我们在实际执行上的变化：只是想让政策更明确）」高风险场景的要求不变，按博文：「we require there to be a qualified "human in the loop" (that is, someone who has the authority to review and, if necessary, change Claude's recommendations), and for the individual affected to be told that AI was used（我们要求有合格的『人在回路』（即有权审查并在必要时修改 Claude 建议的人），且受影响的个人要被告知使用了 AI）」按 TechCrunch（Russell Brandom，发稿时间上午 11:16 PDT）：「Anthropic updated its usage policy on Thursday, codifying new prohibitions on election interference, weapons software, and surveillance.（Anthropic 周四更新了使用政策，把选举干预、武器软件与监控的新禁令成文化。）」以及按 TechCrunch：「Since an update in August, Claude has been trained to end conversations with "persistently harmful or abusive user interactions" — but the new policy explicitly bars users from pursuing those conversations.（自 8 月的一次更新起，Claude 已被训练得会结束带有『持续有害或辱骂性用户互动』的对话——但新政策明确禁止用户去追求这类对话。）」