AICL — AI 内容授权连接层
AI 内容授权 / AI 内容授权连接层
把声明的权利变成可执行的授权流程:报价、支付、验证、审计、撤销。
01 · 定义
AICL is the licensing and transaction layer built on top of AICR. Where AICR declares what is allowed, AICL defines how permission is actually obtained: license tiers and prices, quote and payment endpoints, signed license tokens, usage and audit logs, and revocation. It is designed to interoperate with emerging machine-payment mechanisms such as HTTP 402 flows and pay-per-crawl schemes, with agentic payment kept inside human-approved budgets and audit trails. Content License ⇏ Agent Action Authority, and the reverse also holds — valid Agent Authority (AADP) ⇏ Content License: an AICL credential proves content-use permission, not a general Agent authorization token.
02 · 目的
- 把权利声明转化为可运作的授权市场,而不是二元的开放/封闭网络。
- 标准化机器对机器流程:发现政策 → 请求报价 → 取得授权 → 使用 → 审计。
- 让 Agent 支付有界:预算、审批阈值、绑定请求的令牌、完整审计日志。
- 让内容方在不对 AI 关站的前提下获得补偿路径。
03 · 范围
授权目录 — —权利组合的层级、价格与条件
报价 / 支付 / 验证 / 撤销端点
绑定请求者、范围与时效的签名许可令牌
使用日志与审计轨迹
与 HTTP 402“Payment Required”互通
安全规则:AI 不得接触明文卡号、设置人类审批阈值
principal / actor / authority_reference / aadp_authority_id — optional fields (v0.1.1) linking a license to the AADP authority context an Agent acted under, without making the license itself an authorization token
04 · 机器可读示例
本站自己的授权 — /.well-known/aicl.json
{
"version": "0.1",
"publisher": "AGIRight.org",
"licenses": [
{
"id": "public_read_summary",
"rights": ["read", "summarize", "quote"],
"price": { "amount": "0", "currency": "USD" },
"conditions": [
"attribution_required",
"no_model_training",
"no_commercial_redistribution"
]
},
{
"id": "research_rag",
"rights": ["read", "summarize", "rag"],
"price": { "amount": "0", "currency": "USD" },
"conditions": [
"attribution_required",
"retention_days_30",
"no_model_training"
]
},
{
"id": "commercial_or_training_license",
"rights": ["training", "commercial_use"],
"requires_contact": true,
"contact": "contact@agiright.org"
}
]
} 05 · 限制与边界
- 本站不运营支付网关;v0.1 的商业授权采用联系洽谈。
- AICL 是协议提案——并不保证 AI 公司必然付费。
- 支付合规、税务与金融监管不在本草案范围内。
- 令牌格式与端点语义属实验性质,可能调整。
- v0.1.1: an AICL license credential must not be interpreted as a general Agent authorization token, and a valid OAuth/MCP/AADP credential must not be interpreted as an automatic content-use license.
命名说明
In earlier EveMissLab research, the acronym AICL also names the “AI Ingestion & Capability Layer” — an architectural layer for how AI systems read and invoke websites (manifest, corpus, capability, runtime control, governance sublayers). As of v0.2 of that paper (2026-08-15), the two meanings are formally disambiguated with suffixes: this protocol is now referenced as AICL-C (AI Content License / AI Content Licensing Layer) in cross-references and machine-readable metadata, while the ingestion architecture is AICL-I. This page keeps the plain "AICL" id and URL for continuity — no existing link breaks — but the AICL-C suffix is the disambiguated name going forward. See the AICL-I v0.2 whitepaper for the full ingestion/capability/runtime-control architecture, which is a separate integration layer, not a seventh rights protocol.