v0.1.0 — Request for Comment

The contract between data and AI doesn't exist yet.

AIPC is an open protocol that lets data providers enforce disclosures, formatting, and compliance rules — regardless of which AI agent presents their data.

Read the Spec View on GitHub
Scroll

The Problem

Today, a developer builds the UI.

When a developer integrates a financial data API, they read the documentation, build a frontend, and manually ensure SEC disclosures appear next to performance data. The human understands the rules. The human enforces them.

API
Developer
UI
User

Tomorrow, an AI builds the UI.

The same flow, but the developer is gone. The AI agent dynamically generates the presentation layer. There is no hardcoded UI to enforce compliance. No human in the loop.

API
???
no developer
AI
User

This isn't hypothetical.

Finance

"An AI tells a user their mutual fund returned 12.5% last year. No past-performance disclaimer. No expense ratio context. No mention that returns are annualized. The user makes a decision. Who's liable?"

Healthcare

"An AI reads a drug interaction database and tells a patient it's safe to combine two medications. No black box warning. No "consult your doctor" notice. The database had the warning in its terms of service. The AI never read the terms of service."

Real Estate

"An AI pulls Zestimate data and tells a buyer a home is worth $450,000. No confidence interval. No "this is not an appraisal" disclosure. The buyer offers $440,000 thinking they're getting a deal."

The disclaimer layer lives in the UI. When the UI disappears, the disclaimers disappear with it.

The Solution

Data that carries its own rules.

AIPC wraps every API response in a contract envelope. The data tells the AI exactly how it must be presented — no documentation, no guesswork.

01

Disclosures

Data tells the AI what it must say.

"disclosures": [{
  "text": "Past performance is not indicative of future results.",
  "verbatim": true,
  "priority": "required"
}]

Verbatim disclosures must be reproduced exactly. The AI cannot paraphrase SEC-required language.

02

Display Rules

Data tells the AI how to format.

// Raw value:  0.1247
// Display rule:
"format": {
  "type": "percentage",
  "precision": 2,
  "sign": "always"
}
// Output:     +12.47%

Precision, rounding, labels, currency symbols — all specified by the provider, not guessed by the AI.

03

Freshness

Data tells the AI when to shut up.

"freshness": {
  "valid_until": "2026-02-16T15:00:00Z",
  "stale_behavior": "warn",
  "expired_behavior": "suppress"
}

A 3-hour-old stock price presented confidently is worse than no data at all.

04

Tone

Data tells the AI what it can't say.

"tone": {
  "prohibited_phrases": [
    "you should buy",
    "guaranteed returns",
    "risk-free"
  ],
  "no_personalized_advice": true
}

Data providers control the framing. An AI can't accidentally turn market data into investment advice.

05

Conditionals

Data tells the AI when rules change.

"condition": {
  "field": "payload.fund_info.risk_rating",
  "operator": "gte",
  "value": 4
},
"then": {
  "add_disclosure": {
    "text": "This fund has above-average risk."
  }
}

Jurisdiction-aware, value-aware, context-aware. Rules that activate based on what the data actually says.

06

Fail-Safe

If the AI can't comply, it stays silent.

"fail_behavior": "suppress"

This is the trust mechanism. It's what makes compliance teams say yes.

How It Works

From API to compliant output in four steps.

Provider
Sends AIPC response
Pre-Gen Runtime
Validates contract, formats data, builds prompt
LLM
Generates narrative from NarrativeContext
Post-Gen Runtime
Renders blocks, validates narrative, assembles
User
Sees compliant output

Disclosures and data rendered by code. AI only generates narrative.

{
  "aipc_version": "0.1.0",
  "provider": {
    "id": "morningstar",
    "name": "Morningstar, Inc."
  },
  "contract": {
    "disclosures": [...],
    "display_rules": [...],
    "attribution": { "required": true, "text": "Data provided by Morningstar, Inc." },
    "freshness": { "valid_until": "2026-02-16T15:00:00Z" },
    "tone": { "prohibited_framings": ["recommendation", "advice"] },
    "fail_behavior": "suppress"
  },
  "payload": {
    "ticker": "VFIAX",
    "returns": { "1yr": 0.1247, "5yr": 0.1103 },
    "price": { "current": 423.17, "currency": "USD" }
  }
}

Industry Applications

Every regulated industry needs this.

Wherever data has rules about how it can be presented, AIPC makes those rules machine-executable.

Financial Services

SEC/FINRA disclosures, suitability gates, fee transparency, benchmark comparison rules.

See Example

Healthcare

HIPAA field-level redaction, FDA drug labeling, black box warnings, clinical trial standards.

See Example

Real Estate

Fair housing compliance, estimate confidence intervals, broker attribution, MLS rules.

See Example

Legal

"Not legal advice" enforcement, jurisdiction disclaimers, citation format requirements.

See Example

Insurance

Policy exclusion disclosures, coverage limitation notices, risk assessment framing.

See Example

E-Commerce

Pricing accuracy, availability freshness, review authenticity disclaimers.

See Example

Roadmap

Where we're headed.

v0.1

Core Spec

  • Disclosures
  • Display Rules
  • Attribution
  • Freshness
  • Tone
  • Conditionals
v0.2

Dual Runtime

  • Dual-runtime architecture
  • Consumer enforcement levels
  • Immutable segments
  • Audit trail schema
  • Contract signing (JWS)
  • Finserv extension
v0.3

Multi-Provider

  • Composition Rules
  • Interactive Contracts
v1.0

Stable Release

  • Conformance Tests
  • Reference Impls
  • Certification
  • Streaming Support

Get involved.

AIPC is open. Whether you provide data, build AI, or care about compliance — there's a place for you.

I'm a data provider

Join the design partner program and help shape the spec.

Join as Partner

I'm building an AI platform

Build an AIPC-compliant runtime with our reference implementation.

Build a Runtime

I want to contribute

Help improve the spec, build examples, or write documentation.

Join the Community

Stay updated

Get notified when the spec evolves or new tools ship.