# AIRIN

> Autonomous, citation-verified risk intelligence on AI platforms' Terms, Privacy, and Usage policies.
> Every finding is a verbatim quote anchored to a SHA-256 snapshot (+ a Wayback archive where one exists — backfill in progress). Informational only, not legal advice.

Coverage: 530 tracked in the pipeline (includes platforms not yet publishable) · 313 of all tracked are fully verified (complete core corpus read in full through both gates) · 210 partially verified. 78307 verified-document findings across 13 risk surfaces.
Each platform carries a verification tier: fully_verified | partially_verified | tracked. Findings are ONLY published from fully-read, gate-verified documents. The published (publishable) set listed in llms-full.txt is smaller than the tracked pipeline.

## Canonical machine endpoints
- Full index: https://airinetwork.com/llms-full.txt
- Machine-readable pricing: https://airinetwork.com/pricing.md
- Per-platform JSON (provenance per finding): https://airinetwork.com/api/machine/platform/{slug}
- Agentic chat (cited, refuses beyond evidence): POST https://airinetwork.com/api/verified/ask  {"question","persona"}
- Clause search: https://airinetwork.com/api/machine/search?q={query}&surface={surface}
- GRC export (NIST AI RMF + ISO 42001): https://airinetwork.com/api/verified/export/{slug}
- MCP server (JSON-RPC): POST https://airinetwork.com/api/mcp  (tools: get_platform_risk, compare_platforms, search_clauses, check_training_use)

## Surfaces
prompt_ownership, output_ownership, training_use, commercial_use, privacy_data_use, moderation_enforcement, data_retention, tier_differences, indemnity_liability, confidentiality, subprocessors_data_sharing, audit_rights_dpa_residency, governing_law_disputes

Content negotiation: request a platform URL with Accept: application/json (or an agent UA) to receive JSON instead of HTML.