When your team builds with an AI app builder, three kinds of business data reach the vendor: what people type while building, the records stored in the finished apps, and files they upload. The key questions are whether any of it is used to train AI models, how long it is kept, which outside companies receive it, who can see it, and how you get it back or delete it. Terms vary by vendor and sometimes by plan, so get the answers in writing for the plan you will actually buy.
The first question most owners ask about AI tools is "will it train on our data?" It is a good question, and it is only one of several. When your team builds tools with an AI app builder, your business information travels further than most people realize: into the build conversation, into the apps themselves, into uploaded files, and often through other companies the vendor relies on. Each of those has its own answer about training, storage, access and deletion.
This guide maps where your data actually goes when your team builds with AI, explains the training question and the less obvious ones that matter just as much, and gives you a checklist of questions to put to any vendor before you sign. It is written for owners, IT leads and finance leads, not lawyers, and it does not replace advice from your legal or compliance advisor.
Why this matters now
Your people are already putting business data into AI tools, whether or not you have chosen one. Microsoft and LinkedIn's 2024 Work Trend Index found 78 percent of AI users bring their own AI tools to work. Deloitte UK's 2026 survey found 31 percent of UK workers who use generative AI do so without their employer's knowledge. PagerDuty's 2026 survey found two-thirds of office professionals who use AI at work had used AI tools they believed were not permitted.
That means the realistic choice is not between your data going into AI tools or not. It is between your data going into tools you chose, on terms you read, or into tools you have never heard of, on terms nobody read.
The three kinds of data that reach an AI app builder
| Data | What it is | Example | Why it matters |
|---|---|---|---|
| Build conversation | Everything people type or paste while describing and changing a tool | "Here are 20 of our customers, make the list look like this" | Often stored with the project; may include data people did not mean to share |
| App data | The records stored in and shown by the finished tools | Customer names, job notes, invoices, schedules | Usually the largest and most sensitive set; must be protected from outsiders |
| Uploaded files | Spreadsheets, photos, documents given to the builder or the app | A price list, a photo of a job site, a scanned form | May be stored, processed and kept like other data |
There is often a fourth: usage and activity data, such as who signed in and what they clicked. It is usually less sensitive but still worth asking about.
The build conversation deserves special care
People treat a chat box like a scratch pad. They paste in real customer lists to show the format they want, or a login so the AI can connect to another system. On many platforms that conversation is saved with the project so the AI can keep working on it.
That conversation can be exposed. Lovable says a February 2026 regression exposed the chat history and source code of public projects from February 3 to April 20, 2026. Escape's October 2025 analysis of over 5,600 vibe-coded apps (apps built mostly by describing them to AI) found more than 400 exposed secrets, meaning passwords and keys left where outsiders could read them. Keep sensitive data out of the build conversation entirely. Describe the shape of the data instead: "a customer list with name, phone, address and last visit date."
App data can draw on AI too
Some builders also offer AI features inside the finished apps, such as summarizing notes or drafting replies. Lovable's pricing page, for example, says its credits cover building, hosting with its Cloud, and AI features inside your apps from one balance. When an app uses AI, the records it processes may pass through an AI model provider as well. Ask whether that is the same provider and the same terms as the builder.
The training question
"Training" means using data to improve an AI model, so patterns from your information could influence what the model produces for other people. Most business owners want a clear no for their company data.
Here is what to understand:
- Terms differ by vendor. Some exclude customer data from training by default, some offer an opt-out, and some say little.
- Terms can differ by plan. The same vendor can offer different privacy terms on different plans. The page you read may describe a plan you are not buying.
- Opt-outs can be per person. If an opt-out is a setting each account must change, one person forgetting means their data is not covered.
- Model providers have their own terms. Your vendor may send data to one or more AI model providers. Their terms with your vendor matter too.
- "Improve our services" can mean many things. If the terms say data may be used to improve the service, ask whether that includes training models.
The safest answer is a written commitment, for your plan, that your company's data is not used to train AI models, and that the same applies to every AI provider the vendor uses.
The questions that matter as much as training
How long is data kept?
Ask how long build conversations, app data, files and activity records are kept, and what happens to them when you delete a project or leave the platform.
Who receives it?
Ask for the list of subprocessors, the other companies that handle your data for the vendor, including hosting and AI model providers, and how you are told when the list changes.
Where is it stored?
For some businesses, the country where data is stored matters for legal reasons. Ask, and check with your compliance advisor whether it matters for you.
Who at the vendor can see it?
Ask which of the vendor's staff can access your data, in what circumstances, and whether that access is recorded.
Who can see it at your company?
This is the question people forget. A vendor can have perfect terms and your data can still leak if an app is open to anyone with the link. Our guide to how AI-built apps expose customer records covers this in detail.
How do you get it back, and delete it?
Ask how to export your data in a common format and how to confirm it is deleted when you leave.
A checklist to send every vendor
Copy these questions into an email and ask for written answers for the plan you will buy.
- Is any of our data, including build conversations, app records and files, used to train AI models? Does the answer depend on our plan or on settings each user must change?
- Which outside companies receive our data, including AI model providers, and under what terms?
- How long do you keep build conversations, app data, files and activity records?
- Where is our data stored?
- Which of your staff can access our data, and is that access recorded?
- Can anyone outside our company reach the apps we build, and how is that prevented?
- Do you offer a data processing agreement?
- How do we export all our data, and in what format?
- When we leave, how do we confirm our data is deleted, and how long does it take?
- How will you tell us about a change to these terms, or about an incident affecting our data?
Our AI platform due diligence checklist puts these alongside pricing, access and support questions.
Weak answers and strong answers
Vendors rarely say anything false. The difference is in how specific they are willing to be.
| Question | A weak answer | A strong answer |
|---|---|---|
| Do you train on our data? | "We take privacy seriously" | A written yes or no for your plan, covering every AI provider used |
| Who receives our data? | "Trusted partners" | A named list of subprocessors and how you are told about changes |
| How long is it kept? | "As long as necessary" | Specific periods for conversations, records, files and activity |
| Who can reach our apps? | "Apps are private unless you share them" | Every app lives inside your company and every request checks membership |
| How do we leave? | "Contact support" | A self-serve export in a common format, and confirmed deletion |
Red flags
Slow down and ask more questions if you see any of these:
- Privacy terms that only appear on the highest plan, when you are buying a lower one.
- An opt-out that each employee must switch on individually.
- No list of outside providers, or a refusal to share one.
- Apps that are public by default, or published with one click.
- No way to export your data without contacting sales.
- A history of disputing responsibility for customer data stored in apps built on the platform. Read how the vendor has responded to past incidents, not just its marketing.
Simple rules for your team
Even with the right platform, a few habits prevent most problems.
| Do | Do not |
|---|---|
| Build on the company's approved platform | Build work tools on personal AI accounts |
| Describe the shape of your data in words | Paste real customer lists into the build conversation |
| Use test records while building and testing | Load real records before access checks pass |
| Tell the admin which tools hold sensitive records | Share app links outside the company |
| Ask before putting health or payment data anywhere | Paste passwords or keys into any AI chat |
For anything involving health information, payment cards or other regulated data, read our guide to compliance for AI-built apps and talk to your compliance advisor before building.
Reading vendor terms without a lawyer
You will still want legal advice for anything regulated, but an owner can read the key documents for the important answers in under an hour.
- Find three documents: the privacy policy, the terms of service, and the data processing agreement if there is one.
- Search for "train", "improve", "model" and "machine learning" to find what they say about training.
- Search for "retain", "retention" and "delete" to find how long data is kept.
- Search for "subprocessor" or "third party" to find who else receives your data.
- Check which plan each statement applies to. Look for phrases like "on Business plans" or "for Enterprise customers."
- Write down anything unclear and send it to the vendor as a question.
How FORGE handles it
FORGE was built for companies that put real business data into the tools their teams build, so protecting that data is part of the platform rather than a setting each person has to remember.
- Every app lives inside your company. Every request checks that the person asking belongs to your company. Anyone outside gets nothing back, not even a hint the app exists.
- Only the people you invite. Your admin invites each person by email, every invite works once, and there is no public sign-up.
- The right access for every role. Owners, admins, builders and app users each see and do exactly what their role allows.
- Security checked before it exists. Every piece of every app is checked for quality and security before it is ever saved.
- A complete record. Every step of every build is recorded with who or what did it.
- People leave, tools stay. Removing someone ends their access immediately, and every tool they built stays with your company.
Your IT lead gets a full walkthrough of exactly how FORGE handles your data during your workflow workshop, on your real work. Bring the checklist above. Read more on our security page or talk to us.
Questions people ask
Do AI app builders train AI models on my data?
It depends on the vendor and sometimes on the plan. Some exclude customer data from training by default, some let you opt out, and the terms can differ between plans from the same vendor. Read the terms for the exact plan you will buy, and get the answer in writing.
What happens to what we type while building an app?
The conversation with the builder is usually stored as part of the project so the AI can keep working on it. That means anything pasted into the chat, such as a sample customer list or a password, may be kept. Treat the build conversation as a business record and keep sensitive data out of it.
Can AI app builder chat history be exposed?
It has happened. Lovable says a February 2026 regression exposed the chat history and source code of public projects from February 3 to April 20, 2026, and that it shipped a fix within two hours of the public report.
What is a data processing agreement?
A data processing agreement is a contract that sets out how a vendor handles personal data on your behalf: what it may do with it, how it protects it, which other companies it shares it with, and what happens when the contract ends. Many privacy laws expect one when a vendor handles personal data for you.
What is a subprocessor?
A subprocessor is another company your vendor uses to handle your data, such as a hosting provider or an AI model provider. Your data may pass through several of them. Ask for the list and how you are told when it changes.
What should employees never paste into an AI tool?
Passwords and keys, payment card numbers, health information, government ID numbers, and anything a customer or employee shared in confidence, unless your approved platform and your compliance advisor say that use is covered.
How does FORGE handle our data?
Every FORGE app lives inside your company, and anyone outside it gets nothing back. Only the people you invite can get in, each role sees exactly what it should, and every step is recorded with who or what did it. In your workflow workshop we walk your IT lead through exactly how FORGE handles your data, on your real work.
See what FORGE builds for your team
We onboard every team ourselves, starting with the work that costs you the most time.
Talk to us
