23
module twenty-three
Privacy &Ownership
Where your words actually go.
7 min read
scroll

You just finished writing something raw. A scene, a pitch, an idea — maybe even a piece of your actual life. In some cases, something that felt like a piece of your soul.

You typed it all out, you gave it everything, and now your finger is hovering over enter.

And then the thought hits.

Wait. Where is this thing actually going?

You pull back from the keyboard. You put the phone down. You sit with it — really sit with it — in a way you haven’t before. Not the passing worry you dismissed last time. Something that actually makes you stop.

Good. Stay right there for a second.

Because that pause is exactly where this module begins — and by the time you finish it, you’ll hit enter with something you didn’t have before. Not blind trust. Not paranoia. Just the truth about what happens next.

You clicked "I agree" on a document longer than most short stories. Unlike most of those agreements, this one actually matters.
double-click any line that resonates
// the mechanics

Two Things, Not One

Most people collapse "privacy" into one vague worry. But there are two distinct things happening when your words enter an AI platform, and they work differently.

Storage is whether the platform keeps a record of your conversations. Most do — it's how chat history works. It's generally controllable. You can delete conversations, turn off history, request a data export.

Training is whether your conversations are used to improve future versions of the model. This is the one that matters more for creators. Free tiers typically opt you in by default. Paid tiers give you control. Enterprise accounts almost universally exclude your data entirely. The settings exist on every major platform. Most people never look for them.

Controlling one doesn't mean you've controlled the other.

// platform by platform

The Honest Breakdown

Policies change. What's accurate today may shift with the next update — which is why the move isn't to memorize these details but to know where to find the current ones. That said, here's the honest picture as it stands.

Claude — Anthropic
Training opt-out available — free and paid
Conversations stored temporarily for safety monitoring. Training use is opt-outable in settings — one click, thirty seconds. Projects data has stronger default protections. Anthropic publishes its privacy and usage policies in plain language. So far, they're walking the talk.
The friend who actually reads the fine print and texts you what it says.
ChatGPT — OpenAI
Free: training on by default / Paid: opt-out available
On the free tier, chat history and training are linked — turning off history turns off training. On paid plans, you can disable training independently. The setting lives in Data Controls. Memory features (where it retains information about you across sessions) are separate and independently controllable.
The roommate who eventually learns to ask before borrowing your stuff.
Gemini — Google
Consumer accounts: human review on by default
It's Google. Personal account conversations can be reviewed by human reviewers as part of product improvement — opt-outable, but not by default. Workspace accounts have stronger protections and data is not used for training. If you're using a personal Gmail account, check Activity Controls in your Google account settings.
The roommate who's always listening, even when you think they're not.
Copilot — Microsoft
Commercial plans: strong data protection by default
Copilot through Microsoft 365 commercial plans has enterprise-grade data protection — your prompts and outputs are not used to train foundation models. Consumer Copilot via Bing has less protection. Same interface, very different data handling depending on account type.
The corporation definitely taking notes — but the commercial version keeps them in a locked drawer.
// the real question

Your Work Is Yours. Here's the Nuance.

Across every major platform: you own the output. None of them claim ownership of what you generate. Your work is yours — you can publish it, sell it, put your name on it.

But if your conversations are used for training — even anonymously, even in fragments — they become part of something larger. Not stolen. Not copied. Composted. Your phrasing. Your rhythms. Your creative fingerprints. Still yours. But dissolved into something you can no longer point to.

Nobody's taking your manuscript. But your words entered a system that learns from everything it processes. That's not theft — it's closer to what happens in a compost pile. The original material breaks down. It loses its form. It feeds something larger that you'll never be able to point back to you.

Think of it like a pot of soup that never stops cooking. Every creator who's ever used these platforms added something to it — their phrasing, their rhythms, their instincts about what sounds right. It all dissolves. The soup gets richer. Nobody owns the flavor. Nobody can pull their ingredient back out. Your words are still in there. You just can't find them anymore.

Copyright protects the work. Only you can protect what you shared to build it.
// the practice

Four Moves. That's It.

01 Find your settings once. Every platform has a privacy page. Find the training toggle. Turn it to where you want it. This is a one-time task that most people never do.
02 Match the tier to the work. If you're experimenting, free is fine. If you're working on something that matters — something irreplaceable — work in the environment that offers the most protection. Paid tiers exist partly for this reason.
03 Share what the session needs, not everything you have. Work in sections. You don't need to paste the whole manuscript to get feedback on chapter three. The more sensitive the material, the more deliberately you share it.
04 Keep what you build on your own machine. AI is a workspace, not a storage unit. Everything you create lives on your hardware first. Always.

The next module takes this further — from awareness to active protection. Data exports, timestamped proof of creation, and the architecture for owning your intellectual property in a way that holds up long after any platform's terms of service have changed again.

You own what you made. Know what you shared to make it.
logged
where you are
NowYou know the mechanics — storage vs. training, platform by platform, and what ownership actually means in this space.
NextCopyright and proof — the active layer. Data exports, timestamped IP, and the ledger that makes your ownership undeniable.
ThenThe work is protected. The archive is yours. What was built in the AI stays in your hands — not the platform's.
Continue to Module 24
Copyright & Proof
Your rights. Timestamps. Building the IP ledger.
← HUB COMPOST