Guide
How to use generative AI without your brand assets ending up in training data
Use tools whose terms exclude your content from model training, and keep proprietary or unreleased assets in private models fine-tuned only on material you approve. The account tier and its settings usually decide this more than the tool's name does: the same vendor can train on a free account's content and not on a business account's.
By S77.AI · Updated
What the tools do by default
| Tool and tier | What the vendor says |
|---|---|
| ChatGPT Free, Plus, Pro | Content may be used to train models. You can opt out under Data controls. |
| ChatGPT Business, Enterprise, Edu; OpenAI API | Inputs and outputs are not used to improve models by default. |
| Gemini in Google Workspace | Workspace data is not used to train models outside Workspace without permission. |
| Adobe Firefly | Not trained on user content. Trained on licensed content such as Adobe Stock, and on public domain content. |
| Midjourney | Stealth mode, which controls who can see your images on its website, is limited to Pro and Mega plans. |
Terms change, so check the current ones before uploading anything sensitive. One detail worth knowing: on individual ChatGPT plans, giving feedback on a response can make that whole conversation available for training, even after opting out.
Training is only one of the risks
- Visibility: some tools publish work to a gallery or a shared channel. On Midjourney, creations made in public Discord channels stay visible even with Stealth mode on.
- Personal accounts: business terms only protect work done in the business account. A team member using a personal login falls under different terms.
- Feedback and support: sharing a conversation with the vendor can change how it is handled, as the ChatGPT example above shows.
- Rights: an asset can be safe from training and still not cleared for this use, because of talent, photographer or licensing terms.
When a private model is worth it
A private brand model is fine-tuned on one brand's approved assets. It keeps proprietary material out of public training data by design, not by a setting, and it generates work in that brand's aesthetic instead of a generic AI look, which a general-purpose tool won't do on its own.
It is more setup than a subscription, so match it to the stakes. A business tier may be enough for internal concepts and drafts. Unreleased products, hero campaigns and categories with strict rules on claims or IP are where a private model tends to earn its cost.
What to check before uploading
- Which account tier is this, and what do its current terms say about training?
- Are training and sharing settings controlled by an admin, or left to each user?
- Who can see what gets generated: only you, your workspace, or the public?
- How long does the vendor keep inputs?
- Is this asset cleared for this use, including talent, photography and unreleased product?
- Is there an approved list of assets the team may use?
- Who reviews the output before it is published?
How S77.AI handles it
S77.AI doesn't train public models. When a job calls for it, the studio can fine-tune a private model on a client's approved references, mood boards and asset library. See the AI Content Studio for how that works in production.
Sources
Keep reading
Talk to S77.AI
Have a question about this for your brand? Write to info@s77.ai or see the contact page.