AIRS — AI 权利光谱

01 · 定义

AIRS is the conceptual model underneath AICR and AILP: it holds that AI use of content cannot be governed by a single allow/deny bit. Reading, summarizing, retrieving, transforming, fine-tuning, training, and redistributing are different acts with different depths, uses, and economic weight. AIRS arranges them as a spectrum of levels, each of which can be separately permitted, priced, conditioned, and audited — making rights proportional, use-specific, and trackable. AIRS governs the content-and-learning rights plane only; it does not extend into whether an Agent may act on external systems (AARS) or who authorized that Agent (AADP).

02 · 目的

• 以分级的 AI 访问,取代“开放 vs. 封闭”的假二元。
• 让发布者在浅层最大化开放,同时保留深层权利(训练、再分发)。
• 提供 AICR 声明与 AICL 授权共同引用的词汇。
• 让每一级都可独立审计、独立定价。

03 · 范围

04 · 机器可读示例

机器可读的光谱 — /ai/rights-spectrum.json

{
  "version": "0.1",
  "spectrum": [
    { "level": 0, "name": "No AI Access" },
    { "level": 1, "name": "Read Only" },
    { "level": 2, "name": "Read and Summarize" },
    { "level": 3, "name": "RAG Use" },
    { "level": 4, "name": "Transform" },
    { "level": 5, "name": "Fine-Tuning Use" },
    { "level": 6, "name": "Training Use" },
    { "level": 7, "name": "Commercial Redistribution" }
  ]
}

05 · 限制与边界

• 八个层级是研究模型——真实世界的权利可能需要更细或不同的划分。
• 层级顺序表达的是典型深度与经济权重,不是法律位阶。
• AIRS 描述权利语义,本身不执行也不验证合规。
• v0.1.1: reaching Level 7 on this spectrum still says nothing about Agent action rights or authority — a full Agent workflow may need independent compatibility across AIRS/AILP, AARS, and AADP.