Every AI product now says some version of “your data isn’t used to train our models.” It’s become the privacy equivalent of “all-natural” on a food label — said constantly, explained almost never.
That bothers me, because I’m building a product where the promise really matters. Skybloom is a private space to journal and talk through your whole health with an AI that remembers you. People write things there they wouldn’t say out loud. If “not used for training” is going to be the reason someone trusts us with that, they deserve to know precisely what it means — and what it doesn’t.
So here’s the plain-English version. No legalese, no hand-waving. And because these questions apply to every AI tool you might use, I’ve written it so you can take the same questions to any vendor — including us.
Why private journaling is different with AI in the room
A paper journal has a simple privacy model: whoever holds the notebook reads it. A traditional journaling app adds a company and a server, but your words mostly just sit in storage.
An AI journal is different in kind, not just degree. For the AI to be useful — to notice that your low-energy days keep following late nights, or to remember what you were working through last month — your words have to be processed, not just stored. They pass through AI models. Context gets retrieved and included in prompts. Something about you persists between sessions.
That processing is exactly what makes an AI journal worth using. It’s also exactly where the privacy questions live. So the honest framing isn’t “your data never goes anywhere” — it’s “here is everywhere your data goes, and here is what each party is allowed to do with it.”
What “not used for training” actually means, in plain English
When you write a journal entry or send a chat message in Skybloom, three parties could conceivably touch it: Skybloom, the AI providers that process content on our behalf, and advertisers or data brokers. Let’s take them one at a time.
Skybloom. Your content is encrypted in transit and at rest. We use it to run the product for you — generating your chat responses, your weekly insights, your AI Memory. We do not use your journal or chat content to train Skybloom’s AI models. We never sell your personal information, and we never use it for advertising.
Our AI providers. This is the part most companies gloss over, and it’s the part that matters most. Skybloom doesn’t run every AI model on its own hardware — like most AI products, we use trusted service providers for things like AI chat and voice transcription. When your words pass through those providers, they process them on Skybloom’s behalf, for specified purposes, under agreements that prohibit using your content to train their models. In other words: the “not used for training” promise doesn’t stop at our front door. It follows your words through the whole pipeline.
Advertisers and data brokers. There’s nothing to say here, which is the point. Your words are never sold and never used for ad targeting. There is no version of Skybloom where your 2 a.m. journal entry about stress informs what ads you see anywhere.
One distinction worth making, because it trips people up: processing is not training. When you send a chat message, an AI model obviously has to read it to respond — that’s processing, and no AI product works without it. Training is different: it means your words being folded into a model’s permanent knowledge, potentially influencing what it says to other people forever. Processing happens and ends. Training is what we — and our providers — don’t do with your content.
If you’re evaluating any AI tool for something personal, these are the three questions to ask: Does the company train on my content? Do their AI providers train on my content? Is my content sold or used for ads? A lot of privacy pages answer only the first one loudly and hope you don’t ask the other two.
What the AI remembers — and how you stay in charge of it
“Not used for training” covers what leaves your account. The other half of AI privacy is what stays in it: what the AI remembers about you.
Skybloom uses a single AI Memory shared across chat and journaling. It holds health context you’ve shared — the things worth remembering so you don’t have to re-explain yourself every session. And it’s built to be inspected, not hidden:
- You can read it. AI Memory isn’t a black box. You can open it and see exactly what the AI is carrying forward about you.
- You can edit it. If something in there is wrong, outdated, or just something you’d rather the AI not hold onto, you can change or remove it.
- You can see when journal context was accessed. Skybloom shows you when your journal entries were retrieved as context, so “the AI read my journal” is a visible event, not a vague possibility.
- You can opt entries out entirely. Mark a journal entry Extra Private and it won’t be retrieved for AI context at all. It’s yours to write; the AI simply doesn’t get to use it.
- You can leave with your data. Export a copy of everything, or delete your account.
The payoff of this memory, used over time, is Skybloom’s weekly insights: with consistent journaling and chats, weekly themes can surface patterns across your reflections — the kind of dot-connecting that’s hard to do yourself across scattered entries. The reason I can offer that feature without flinching is everything above: the memory doing the connecting is one you can read, edit, and fence off.
Who Skybloom is for — and who it isn’t for
I’d rather be clear than persuasive here.
Skybloom is for you if you want a private place to reflect on your whole health — journaling (typed, prompted, or voice), health conversations, or both — and you like the idea of an AI that remembers your context and helps you notice patterns over time. It’s especially for you if you’ve hesitated to put personal things into general-purpose chatbots because you weren’t sure where your words would end up.
Skybloom is not for you if you’re looking for medical answers. This matters most with topics like sleep, nutrition, supplements, or medications — exactly the things people naturally want to think through. Skybloom Chat is a good place to think out loud about a stretch of bad sleep and organize the questions you want to bring to your doctor. It is not a place to get a diagnosis, dosing guidance, or treatment advice, and it won’t pretend to be.
Skybloom is a wellbeing and reflection tool — not medical care, professional advice, or an emergency service. If you’re dealing with a medical issue, the right destination is a professional; Skybloom’s job is to help you arrive at that conversation with clearer questions.
The short version
Your journal and chat content in Skybloom: encrypted in transit and at rest, never sold, never used for advertising, and never used to train Skybloom’s AI models or the models of the AI providers that process content on our behalf. What the AI remembers about you lives in an AI Memory you can read and edit, with journal access you can see, entries you can mark Extra Private, and an export/delete door that’s always open.
That’s what “not used to train AI” means here — spelled out, because a promise you can’t inspect isn’t much of a promise.
If that’s the kind of private space you’ve been looking for, Skybloom is free to start: 5 chat messages a day, 3 journal entries a week, and one guided program, with no credit card required — and the free plan stays free unless you choose to upgrade.