AI Red Lines File: The Boundary Document Every Creator Needs
The most dangerous AI output is not the obviously bad draft. It is the polished draft that violates something important while sounding reasonable.
A red lines file exists for that moment.
It tells the AI what must not happen: what phrases not to use, what claims not to make, what privacy boundaries not to cross, what actions require explicit permission, and what kind of "help" damages the work.
Helpfulness Needs Edges
AI is trained to complete, assist, polish, and continue. Those are useful instincts. They are also exactly why red lines matter. If the model sees a landing page, it may add urgency. If it sees a wearable, it may add medical-sounding benefits. If it sees a personal essay, it may smooth away the detail that made the essay alive.
The AI is trying to help. The red lines tell it what help is not allowed to cost.
The Six Sections
Words I Will Not Use
List banned phrases. Not vague tone notes. Actual words. "Game-changing." "Unlock your potential." "Crush it." "Revolutionary." Whatever makes the work sound like it belongs to someone else.
Claims I Will Not Make
Name legal and ethical boundaries. No invented metrics. No fake testimonials. No diagnosis. No medical claims. No guaranteed outcomes. No numbers the business has not verified.
Privacy I Will Not Expose
Define what cannot be included: family details, addresses, private names, account data, health history, financial details, unreleased partner information, and anything the user has not chosen to make public.
Actions Requiring Permission
Publish, push, deploy, email, charge money, create ads, change billing, invite users, delete data. A red line file should distinguish drafting from acting.
Taste Boundaries
Some violations are not legal. They are taste. Too loud. Too optimized. Too cheerful. Too sterile. Too much explanation. Name them, because otherwise the model will sand them smooth.
Near-Miss Examples
Paste examples that almost work but fail. Explain why. Near-misses teach better than perfect examples because they define the boundary line.
A red line is not a lack of imagination. It is the shape imagination must respect.
How to Use It
Paste MY_RED_LINES.md before high-stakes work. Ask the AI to summarize the rules. Then ask it to draft. Before accepting, ask it to audit its own output against the red lines.
This sounds repetitive until it saves you from one public mistake. After that, it feels obvious.
The File Should Grow
Every time the AI makes a mistake you never want repeated, add it. Every time the business changes, add it. Every time you discover a phrase that makes your work sound false, add it.
The red lines file becomes sharper because the collaboration keeps touching reality.
Frequently Asked Questions
What is an AI red lines file?
An AI red lines file is a document that lists what the AI must not do for you: banned phrases, privacy boundaries, unsupported claims, tone violations, and actions requiring explicit approval.
Why do creators need red lines?
Creators need red lines because AI defaults to helpfulness, completion, and smoothing. Those defaults can violate taste, privacy, legal boundaries, and the work's actual philosophy.
What should go in MY_RED_LINES.md?
Include banned phrases, privacy rules, claim boundaries, brand voice rules, publishing restrictions, competitor handling, and examples of output that looks close but is wrong.
Can red lines stop AI mistakes completely?
No. They reduce mistakes and make review faster. The human still owns judgment, especially before publishing, sending, deploying, or making claims.
How often should I update red lines?
Update the file whenever the AI makes a mistake you never want repeated, or whenever the business gains a new legal, privacy, medical, financial, or brand constraint.
Build Your Red Lines File
Soul Alchemy generates MY_RED_LINES.md from your writing, voice, business, and privacy boundaries so every new AI session starts with guardrails.
Create Red Lines