---
title: "Honest privacy limits"
lesson_id: "05"
---

# Honest privacy limits

- **Lesson ID:** 05
- **Goal:** State honest privacy limits: default no transcripts, consent for text, synthetic leak risk, LoRA memorization, and bake-as-release.
- **Human lesson:** [05-privacy-limits.html](05-privacy-limits.html)

## Prerequisites

- Lesson 04, or a baked .aimodel path.

## Inputs, outputs, and artifacts

- **Inputs:** The full loop.
- **Outputs:** A six-item limit list.
- **Artifacts:** Progressive `.flow` diagrams and an ordered list.

## Agent build steps

1. Repeat the default no-transcript rule.
2. List the six README limits.
3. Prefer prompt rollback before a bake when tone is the bug.
4. Link prompt registry.

## Constraints

Keep the language simple, friendly, and literal. iOS 27 and macOS 27 only. No quizzes, Mermaid, or em dashes. Do not invent API names. Mark sketches as sketches and tell the reader to confirm them in the Xcode 27 SDK.

## Key concepts

- Default telemetry has no transcripts.
- Text leaves the device only after a clear yes to systems you name.
- A LoRA can memorize a rare string. Do not train on secrets.
- A new `.aimodel` is an app release, not a silent system model update.

## Takeaways

- Signals first, synthetic rows with review, LoRA on a Mac, bake to Core AI.
- Drop rows when the seed is too close to a user.
- Fix tool paths and evals before training when the failure is not model knowledge.

## Acceptance checks

- Six limits are listed.
- Public trainer upload is not the default.
- Sibling links, progressive `.flow` diagrams, and an Agent brief link are present.