Can you send client code to AI?

# Client-code permission review

This pack is a review aid, not an NDA, legal opinion, consent or contract amendment. It sends no message and makes no API call. Do not put actual client source code or credentials into the training files.

Use permission-review.md with the responsible legal/commercial owner and the client’s authorised process. Record exact scope and evidence in your approved system. The toy function and cases were written independently for this guide and contain no client information. They illustrate a clean reproduction, not a guarantee that modified client code is free of restrictions.

## Guide and sources

Canonical guide: https://aona.ai/resources/guides/client-source-code-ai-nda/
Source review: 2026-09-21
- USPTO: Guidance on use of AI-based tools: https://www.govinfo.gov/content/pkg/FR-2024-04-11/html/2024-07629.htm

fictional-permission-case.md is a completed teaching example: the fictional authority withholds original client code and allows only the independent toy reproduction within its stated example scope. It is separate from the blank actual-client review and grants no real permission.

# Proposed client-code disclosure

Client/reference: ____________________
Governing agreement and relevant review owner: ____________________
Material proposed, including derived/automatic context: ____________________
Service, account and additional recipients: ____________________
Purpose and expected benefit: ____________________
Provider handling evidence for the client to consider: ____________________
Who is authorised to decide under the client process: ____________________
Decision/evidence reference: NOT OBTAINED
Limits and recheck triggers: ____________________

This file grants no permission and does not amend an agreement.

# Independently written synthetic example. No I/O or client code.
def require_quantity(payload):
    if "quantity" not in payload:
        raise ValueError("quantity is required")
    return payload["quantity"]

{
  "notice": "INDEPENDENT SYNTHETIC EXAMPLES ONLY",
  "cases": [
    {
      "input": {
        "quantity": 2
      },
      "expected_return": 2
    },
    {
      "input": {},
      "expected_exception": "ValueError",
      "expected_message": "quantity is required"
    }
  ],
  "client_material_used": false,
  "provider_calls_performed": false
}

# Clean reproduction review

General programming question: ____________________
Independently written example: ____________________
Invented names/values and minimal structure: ____________________
Client-specific algorithm, comments or business rules excluded: ____________________
Does any derived information still require permission? ____________________
Reviewer and decision: NOT YET REVIEWED

A renamed copy is not automatically an independent example.

# Worked fictional client-permission decision

SYNTHETIC SCENARIO. This is not consent, legal advice or a contract amendment.

Request: a software agency wants an AI explanation of a quantity-check failure in SYNTHETIC_CLIENT_PORTAL.
Material proposed: the original fictional client function and nearby context. No actual client function is included in this pack.
Recipient path: SYNTHETIC_AI_SERVICE in SYNTHETIC_WORKSPACE, then a fictional model provider. These are labels, not real services.
Authority in this exercise: only the fictional client contract owner may decide disclosure scope; the project manager cannot extend it.
Decision: original client code is not authorised for this use. Use only clean-example.py, independently written for the guide, with synthetic-cases.json.
Limits: no client names, algorithms, comments, surrounding files or production values may be added under this illustrative decision.
Recheck: any change to material, purpose, account or recipients requires a new authorised review.
Outcome: a useful general programming question can proceed with the independent synthetic example while the client-code request stays unresolved.