Incident #A7EF80 · 09 SEP 2026major
Lovable rewrites privacy policy: AI training on your content by default, ad sharing added
The policy was fully rewritten. It now states that Customer Content (prompts, code) and Usage Data are used to train Lovable's AI models by default, with an opt-out on any plan that is not retroactive. It also adds sharing of pseudonymized identifiers with ad platforms (Meta, Google), session recording, and an account-claim process for organizations.
Taken from you
- Prior promise that project artefacts are never used to train general-purpose models without permission is gone; training on your content is now default
- Opt-out of training does not remove content from datasets or models already trained
- Pseudonymized identifiers may be shared with ad platforms such as Meta and Google, treated as a 'sale' or 'sharing' in the US
- Sampled session recordings of your interactions with the product, where consent applies
- Employer or organization can claim your account if you signed up with a work email and you do not choose
- Trained staff may review your content for model quality
- Specific security commitments (72-hour breach notice, SOC 2 audit, one-year log retention) replaced by general language
- Delaware courts clause and the 30-day request response period are no longer stated
Won back
- Opt-out of AI training is now a self-serve setting in your account, free on any plan. Before, you had to email Lovable or upgrade to a Business plan.
- Added Brazil LGPD rights and a named Swedish controller and lead supervisory authority
- Clearer rights to human review of significant automated decisions
- Notice before domain-claim account transfer, with the option to change your email to keep the account personal
Evidence · 4 of 4 quotes matched in source
We use Customer Content and Usage Data — which may include Personal Data — to train, develop, fine-tune, and improve our AI models and AI-powered features+ inserted · ✓ matched in source
It does not retract content from training datasets assembled, or models trained, before you opted out.+ inserted · ✓ matched in source
we share limited Personal Data — pseudonymized identifiers, not the contents of your projects — with advertising platforms such as Meta and Google+ inserted · ✓ matched in source
Archived version ↗ · 7,290 words changed