In partnership with

Ever built a little app with AI and felt like a developer for an afternoon?

A waitlist, a booking page, a client portal. You described it to Lovable, and people started signing up.

The question I'd ask today is simple: who else can see those sign-ups?

A security firm in Hamburg called Crissy Field looked at Lovable apps built on the Supabase database, together with the German paper Handelsblatt, which reported it on November 2.

And on a lot of them, personal data loaded for anyone, without logging in.

A forum with full names and birthdays. Ticket buyers with phone numbers and addresses. Waiting lists and newsletter sign-ups. And on one service, passport details.

(The firm is careful here: it says its numbers are not a confirmed leak rate, and its sample is not all Lovable apps. It also didn't publish how it got in, which is right.)

The report also points out that Lovable's terms put the security of your app and your data on you, the customer.

I think it's true that anyone can build an app now. But keeping its data safe is the part that isn't on the homepage.

Right after this: the free scan that's already inside Lovable, the two-minute check you can do yourself, a prompt for a plain-English review, and when to get a human.

Now a quick word from today's sponsor.

Master Claude AI (Free Guide)

The professionals pulling ahead aren't working more. They're using Claude.

Our free guide will show you how to:

  • Configure Claude to be the perfect assistant

  • Master AI-powered content creation

  • Transform complex data into actionable strategies

  • Harness Claude’s full potential

Transform your workflow with AI and stay ahead of the curve with this comprehensive guide to using Claude at work.

Right. Back to it.

Run the free scan first

Lovable has a security check built in, and you've probably only seen it as one line in the publish window.

  1. Open your project.

  2. In the project toolbar, go to More → Security.

  3. Click Run deep scan.

It reads all your code and your database rules, and it usually takes 3 to 13 minutes.

Both the quick scan and the deep scan are free, and they do not use credits.

So start with anything marked critical. Many findings come with a fix you can ask Lovable to make, and you review the change.

(One detail I didn't know: findings don't stop you from publishing by default. The publish window shows them, and you can click straight past.)

And if you built on Supabase yourself, open your project there and check Advisors → Security Advisor. Same idea, different screen.

The two-minute check

Open your app in a private browser window, where you are not logged in.

Then click through every page.

If you see anyone's name, email, order or message there, so can a stranger.

And if it looks clean, that does not prove it is. The data in the report came out through requests a normal page never shows you. That's why I'd trust the scan over this check.

Ask for a plain-English review

Paste this into your project chat:

Review my app's security. Don't change anything yet.

1. List every table that stores personal data: names, emails, phone numbers, addresses, payments, documents.
2. For each one, tell me who can read it and who can change it: a visitor who isn't logged in, a logged-in user, and a different logged-in user.
3. Flag anything a visitor who isn't logged in can read that they shouldn't.
4. Explain each problem in plain English, suggest the fix, and wait for my OK before you change anything.

My app is for [what it does]. Only [who] should see [what data].

That last line matters, because the AI can't know who is supposed to see what unless you tell it.

(Heads up: a security review in the chat uses credits. The deep scan does not.)

When to get a human

Honestly, if your app holds payments, ID documents, health details or anything about other people's customers, a scan is not enough.

The report's own advice is to have someone check who can access what before real customer data goes in, and again after every big change.

And until then, I'd keep real personal data out of it. Test sign-ups with your own email are fine, though.

Make it stick

Run the deep scan once today. You don't need to fix everything at once, but I would fix anything critical before the next sign-up comes in.

Then run it again every time you add a feature that stores something new.

That's the whole habit.

Building is the easy part now.

You can get an app, a page or a whole funnel out of AI in an afternoon.

Then you spend the week fixing it and explaining again what you meant.

Overpowered is the complete AI system I run my business on. It knows you, works with any AI and gets better the more you use it.

It's $29 a month.

Talk tomorrow,
Zephyr

P.S. Did the scan find anything on yours? Hit reply and tell me what it flagged. I want to know what comes up most. I won't share names.