Lovable is turning AI training
on by default.
Lovable, the "vibe coding" tool that generates entire web apps from a prompt, is flipping the default on whether Free and Pro usage data trains its AI models. The cutoff is September 9 — miss it, and your confidential prototype becomes training-data material by default.
The end of "no clear default"
Lovable is one of the leading "vibe coding" tools that generate a working web app from a text prompt. Until now, there was no clear default on whether the code and prompts you generated would train Lovable's AI models, regardless of plan. From September 9, 2026, Free and Pro usage data will be used to train Lovable's AI models by default. The change is spelled out in Section 5 of the updated Privacy Policy and in the newly published support document, "Manage training data and privacy".
The definition of "customer data" subject to training is specific: prompts (including attached images and files), generated code, project files, hosted applications, configurations, and generated outputs — all of it is in scope. Account and billing details are excluded, and so is any data that your app's end users submit directly to your app — that stays in your own project's database and storage.
What's in scope, what isn't
The default flips entirely depending on whether you're on Free/Pro or Business/Enterprise.
Workspace data on Business and Enterprise plans is excluded from training by default. That treatment is set separately by each organization's Data Processing Agreement (DPA) — it's a contractual term, not an optional setting. Free and Pro, by contrast, have exactly one control point: your account settings. Miss it, and your data stays in scope by design.
How the two tiers differ
| Free · Pro | Business · Enterprise |
|---|---|
| Training on by default from September 9 | Excluded from training by default |
| Opt out anytime in Account settings | Protected by a Data Processing Agreement |
| In scope: prompts, code, generated outputs, etc. | Workspace data not used for training |
| End-user submitted data excluded (both) | Same (both) |
Three steps to opt out right now
You can change the setting anytime. Turn it off before September 9 and your data is never used for training.
Open Account settings
Log in to Lovable and open the "AI model training" section under Account settings.
Switch the toggle off
Turn off "Use my Lovable content for model training." The change takes effect immediately.
Confirm what it covers
Opt out before September 9 and your data is never used; opt out after that date and your data is excluded going forward. The official FAQ describes the same behavior.
"If you don't want your data used for training, notice the setting and change it yourself."
That is fast becoming the default across AI dev tools.
What it means for each role
For engineers
If you're building solo projects or client work on Free or Pro, check your opt-out setting before September 9. NDA-bound work and confidential prototypes are in scope — code included.
For leadership and IT
Audit whether any team or contractor is running Lovable on Free or Pro. For sensitive work, upgrading to Business or Enterprise — excluded by default — is a real option worth considering.
For PMs and product
Unreleased specs and competitive analysis you're prototyping with also fall under "customer data." Add the setting check to your team's checklist.
What happens next
This kind of "opt-out by default" training policy appears to be spreading across the AI dev tool industry. Services whose main customers are solo builders and startups have a strong incentive to feed free and low-tier usage data back into model improvement. Even within "vibe coding" tools, some already let you flip on a one-click training stop — the default design is starting to diverge by vendor. More coding tools are likely to follow with similar default changes.
The practical response is simple: (1) check where your toggle currently sits in Account settings, (2) switch it off before September 9 if needed, and (3) notify any team working on NDA-bound or confidential projects about this change. Those three steps cover the near-term action.
This isn't all upside
Designing "opt-out" as the default invites criticism. Most users don't proactively check settings pages, so it seems likely that a majority end up in scope for training simply by not noticing. It's also worth discounting for the fact that this design can function as a nudge toward upgrading to Business or Enterprise.
Public information doesn't clearly address how far back existing project data is affected once September 9 passes. Lovable states that opting out excludes your data going forward, but the treatment of data already accumulated before the change isn't explicitly addressed — that likely needs more clarification.