Ask HN: Software Licenses that prevent LLMs from training on open source?

hn Ask HN · 原文链接 ↗ · 29 分

I'm planning to release an open source software library soon and was wondering - are there any licenses that would prevent LLMs from ingesting it into their training? Do any of the labs say that they don't ingest software code that uses certain licenses? I realize this would be something that would be nearly impossible to discover and they could always say that they don't but they end up ingesting anyway.

五维评分

AI 分析

痛点：开源作者想阻止自己的代码被 LLM 抓取训练，但现有开源许可证普遍不禁止 AI 训练，且缺乏可验证的检测与执行手段。

目标用户：准备发布开源库的独立开发者、对 AI 抓取敏感的开源项目维护者、需要合规声明的商业公司开源团队。

证据：are there any licenses that would prevent LLMs from ingesting it into their training? Do any of the labs say that they don't ingest software code that uses certain licenses? I realize this would be something that would be nearly impossible to discover and they could always say that they don't but they end up ingesting anyway.

变现：免费提供许可证生成器获客，付费售卖「AI 训练 opt-out 合规包」：自定义许可证模板 + robots.txt/ai.txt 配置 + 代码水印与训练数据溯源检测订阅（按仓库月费）。

MVP：两周内做一个「AI-Opt-Out License Generator」：输入项目信息，一键生成附带禁止 AI 训练条款的 LICENSE、ai.txt/robots.txt、README 声明片段和 GitHub 仓库自动提交 Action；先靠免费工具收集邮箱，再卖溯源检测报告。

竞品：Hippocratic License、Do No Harm License、BSL / Elastic License、Spawning 的 Have I Been Trained / Do Not Train 注册表、ai.txt 提案、Creative Commons 关于 AI 训练的讨论稿

摘要：HN 上有人提问：是否存在能阻止 LLM 拿自己代码训练的许可证，并质疑即便有条款也几乎无法发现和执行的违规行为。帖子仅 7 分 5 评论，说明这是萌芽期的真实焦虑但尚未形成明确付费市场，且已有 Hippocratic License、Do Not Train 注册表等竞品。独立开发者可做低成本的许可证生成+声明工具切入，但变现路径需要靠溯源检测等增值服务补足。