ARTICLE
AI Safety for Small Business: How to Protect Your IP Without Paying a Cent
July 10, 2026
Your business is yours. Your offerings, your pricing logic, your client process, the way you solve problems that your competitors haven't cracked yet. That's worth protecting. So when someone says, "be careful what you put into AI." The obvious next question is, “Careful?…How? ”What actually happens to your information the moment you hit enter? Does the AI remember it? Is it quietly being used to train the next version of the model? The one your competitor might be using next month? Is a real person somewhere reading it? Where's the actual line between "safe to share" and "now baked into someone else's product"? Here's the short answer: by default, in most tools, yes — it can be remembered, used to train future models, and read by a human reviewer. But "by default" is the important word. Every major AI tool gives you a way to switch each of those off. Most people just don't know where the switch is. So here it is, tool by tool, for the popular AI platforms you're most likely already using.
ChatGPT (Free, Plus, or Pro)
Go to Settings, then Data Controls, and turn off "Improve the model for everyone." This is the setting that decides whether your conversations get used to train future models. Bear in mind that OpenAI still retains conversations for up to 30 days for safety and abuse monitoring, even with this switched off. That's standard across most AI tools, not a ChatGPT-specific gap. This setting doesn't switch off file uploads either. You can keep attaching documents as normal.
Claude (Free, Pro, or Max)
This one's the opposite of what you'd expect. Claude trains on your conversations by default, but the opt-out is one toggle away in Settings, then Privacy ("Help improve Claude"). It doesn't cost you the ability to upload files. Turn it off once and it applies to everything going forward.
M365 Copilot (Business Basic or Standard)
You likely already have Microsoft 365 Copilot Chat included at no extra cost. Sign in with your work account, and enterprise data protection switches on automatically. You'll see a green shield in the interface confirming it. Your prompts and files aren't used to train the underlying AI models, and no human reviewer is reading them. If you're already paying for Microsoft 365, it’s the lowest-effort AI safety option available. Nothing to buy. Nothing to configure. Simply log in and go.
Google Gemini (Free, Plus, Pro, or Ultra)
Gemini works a little differently to the others, and it's important to understand why. By default, Google can use your conversations to improve its models, and a portion of those conversations are read by human reviewers to judge response quality. Google itself warns users not to type anything into Gemini that they wouldn't want a reviewer to see. You have two free ways to turn this off:
Option 1: Turn off Gemini Apps Activity
Go to myactivity.google.com/product/gemini, find Gemini Apps Activity, and turn it off. Future conversations won't be sent for human review or used for training. The trade-off: without saved activity, Gemini won't hold onto any long-running chat history between sessions, so it's a genuine trade-off between privacy and a persistent memory of past chats.
Option 2: Use Temporary Chat
If you want to keep your everyday chat history active but need a private space for one sensitive task, like drafting a pitch deck or reviewing a contract, open a Temporary Chat from the menu next to "New chat." Nothing in it is used for training or human review, and once you leave it, it's gone. Copy anything valuable into a document before you close it. Also worth checking if you use Gemini with a personal Gmail account: Google has, by default, connected Gemini to your Gmail, Drive, and Docs for many users. Look in Gemini's settings for Personal Intelligence or Connected Apps, and switch off anything you don't want Gemini reading, particularly if you're drafting financials or strategy documents in your personal Drive.
If you're uncomfortable changing your AI settings
There's an alternative. An easy, low-tech option that works on any AI tool, right now, with zero setup. Don't tell it who you really are. Save this on your desktop and paste it into any new chat before you start:
COPY THIS PROMPT
“Act as an expert startup consultant. To protect privacy, I have anonymised this data. Treat ‘Company X’ as my business, ‘Product Y’ as our offer, and ‘Competitor Z’ as our main rival. Do not ask for real names. Here’s the project text: [insert here]”
Sounds almost too simple? It works because the AI doesn't need your real names to give you useful thinking. It needs the actual problem, which you're still giving it in full.
The Setting You Can't Toggle
Settings only get you so far. Even with training switched off, most tools still store your conversation briefly for safety review, and no free tool gives you an ironclad guarantee. So the second half of AI IP protection is judgment, not just settings: keep the strategic thinking, your pricing model, your positioning, your unreleased plans, in AI tools where you've confirmed LLM training and human review are switched off, and think twice before pasting anything you'd be uncomfortable seeing turn up in someone else's answer six months from now.
Pick the option that matches your setup, make the changes you need, and you're back to work.
About the Author
Stephanie Silvester is the founder of The UnConsultants, The Agentic CMO Lab and Uncomplicated AI for Business. After 30 years leading marketing inside global organisations, she now helps small business owners make sense of AI without the jargon - as someone who's made the calls and knows what actually matters when you're running the whole show yourself.
The information on this website is general in nature and does not constitute professional advice. All rights reserved.
Other names and brands may be claimed as the property of others.
© 2026 The UnConsultants Pty Ltd
ABN 87 689 593 333