AILP — AI Learning Permission Protocol

AI Learning Permission Protocol

Can AI learn from this? To what depth, for which uses, with what obligations?

01 · Definition

AILP operationalizes the AIRS spectrum for the specific question of learning. It distinguishes acts that binary crawler rules collapse together: being read is not being learned from; being retrieved is not being trained on; fine-tuning is not distillation; verbatim memorization is different from statistical internalization. AILP lets a publisher declare, in a single machine-readable file, per-dimension permissions — access, indexing, inference input, embedding, training, fine-tuning, distillation, memory, output, attribution and compensation — each as allowed, denied, or license-required.

02 · Purpose

Answer "may AI learn from this?" with dimension-level precision instead of a crawl bit.
Separate learning permission (internalizing into models) from access permission (fetching pages).
Support compensation models: free, non-commercial, license-required, revenue-share.
Ensure openness is machine-executable — so goodwill is not discarded as legal uncertainty.

03 · Scope

access / indexing — fetching and semantic indexing

inference_input — use as context at inference time (RAG)

embedding — vectorization and retrieval indexes

training / fine_tuning / distillation — model internalization tiers

verbatim_memory — whether exact reproduction is permitted

attribution / compensation — citation duties and payment terms

04 · Machine-readable example

A learning-permission declaration — /ai/rights-spectrum.json (AILP profile)

/ai/rights-spectrum.json

{
  "version": "0.1",
  "protocol": "AILP",
  "publisher": "example.org",
  "default": {
    "access": "allowed",
    "indexing": "allowed",
    "inference_input": "allowed",
    "embedding": "allowed",
    "training": "license_required",
    "fine_tuning": "license_required",
    "distillation": "denied",
    "verbatim_memory": "denied",
    "attribution": "required",
    "compensation": "contact"
  },
  "contact": "licensing@example.org"
}

05 · Limitations

Learning-permission semantics are the hardest to verify technically — AILP declares intent, it cannot yet prove compliance.
The dimension list is a draft; boundaries between training, fine-tuning, and distillation are still debated.
Legal recognition of learning permissions varies by jurisdiction and is unresolved.