cd /news/ai-safety/a-proposal-to-prevent-mass-security-… · home › topics › ai-safety › article
[ARTICLE · art-171955] src=skape.run ↗ pub= topic=ai-safety verified=true sentiment=↓ negative

A proposal to prevent mass security incidents like Lovable's

A security incident discovered by Crissy Field GmbH left the databases of 24,664 websites built on Lovable publicly accessible, exposing data including passport numbers, birth dates, invoices, contracts, names and email addresses. Lovable responded that users must monitor security scans and fix flagged issues, prompting Skape founder — whose app platform is secure by default — to argue that platforms should build security in rather than scan for it, with apps private by default, platform-handled login and per-role query permissions instead of table-by-table Row Level Security.

recommendation

Treat agent-built apps as untrusted: make apps and data private by default, handle login and data routing in the platform, and grant each role only specific allowed queries instead of relying on users to fix scan findings.

more in playbook: Securing agents and AI tools
read5 min views4 publishedNov 18, 2026
A proposal to prevent mass security incidents like Lovable's
Image: source

I'm building Skape, an app platform that's secure by default. This essay argues that platforms like Lovable need security built in rather than scanned for, and shows how Skape does it.

Lovable is a service that allows anyone to build websites with AI. In a recent Lovable security incident discovered by Crissy Field GmbH, 24,664 websites had their databases publicly accessible. What was exposed varied by site, but some of the information exposed included passport numbers, birth dates, invoices, contracts, names, and email addresses.

This is emblematic of a larger problem with the overall idea that we can give tools designed for engineers to non-coders and expect good results when they code with agents. In 2026, if something is accessible on the internet, you cannot make any security mistakes. The challenge for coding platforms is that in order for apps to be really useful, users want them to be accessible everywhere and to everyone.

Lovable's response to the security issues raised in the incident was that users need to be paying attention to security scans and fixing the issues raised. Security issues have gotten more and more complex over time, and it is a stretch for a platform to expect that non-developers will spend time worrying about whether their app is susceptible to subtle but dangerous security attacks or that they will understand the tradeoffs of the mitigations that agents will propose. When we have more people building, with less security knowledge, we increase the likelihood of attacks succeeding.

While I appreciate that Lovable is doing their best, instead of trying to scan for security issues (which is a game of whack-a-mole) we need to give non-technical users a platform where security is the default. Folks deserve a place where they can be creative and build without worrying about subtle security issues.

Introducing Skape #

Like Lovable, Skape is an app platform where you can quickly build apps. However, Skape has security built into the platform. Apps are treated as untrusted by the platform, with users granting permissions for what apps can do.

Skape has multiple layers of security that help prevent your data getting leaked:

  • Apps are private by default - you can choose to share your app with specific people, or publish it publicly, but until then no one can see it.
  • Data is private by default - each user within an app has their own private data. When you want to collaborate with someone, you invite them to your copy of the app (again, using skape's built in functionality, not something your agent codes).
  • Apps never handle security functionality - login and sharing data with other users all happens in the platform. This doesn't mean "your app shows its own UI and then feeds username/password to a login service" - I mean that Skape's UI handles login, and then while you use your app, Skape routes your data requests to your database. Your agent can't make a mistake in code it never wrote.
  • Each user role can only run specific queries - Row Level Security systems (like the one Lovable uses through Supabase) require you to use a complex configuration system of policies that must be turned on table by table - Skape's system of "just give each role a set of queries they are allowed to run" is much simpler to reason about, and is a better fit for how apps are actually built.

Would Skape have prevented the Lovable security incident? #

None of this data would have been reachable by a public internet security scan if the apps had been built on Skape. Skape's security measures mean that app creators get secure defaults - most users' data never would have been accessible at all because only invited users can see their data. The caveat discussed below is that some of the apps couldn't have been built on Skape.

In the case of apps where different users have different permissions and some data is meant to be hidden from others, business logic will always determine who can see what, so humans and agents will always have to be involved in that code (regardless of the platform). But the fact that users can only run defined queries means that we've massively simplified the review.

Tradeoffs & Caveats #

Skape is built to support collaborative apps where a single user works with data or small to medium groups of users work together with data. It can't support public forum style apps or apps where anyone on the internet can submit info for a website owner to review. For published apps, anyone can open the app - but since your copy of the app is private to you and anyone you invite, anonymous users can't see it, and new users get their own private space. For many of the affected apps, I suspect the creator didn't need to have everyone working with the same database and exposing their users' data wasn't worth it.

Additionally, moving security into the platform means that if there is a security issue in the platform, all apps would be affected. But would you rather have a security feature like sharing implemented by software engineers focused on building just that feature vs thousands of non-developers who are trying to quickly get a feature out the door on their vibe coded app? I think that concentrating the risk in the platform is the right tradeoff for this environment.

Skape makes it more fun to build apps #

What's exciting to me is that since I don't have to spend time worrying about security in the apps I'm building on the platform, I find it very freeing - I can just focus on having fun building the apps.

Skape makes building apps startlingly simple - it only takes a minute to connect your agent and get started creating an app.

Try it out now #

Get started building right now - not sure what to build? Is there an app you use every day that you'd love to slightly tweak? Try building your own version on Skape.

── more in #ai-safety 4 stories · sorted by recency
── more on @lovable 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/a-proposal-to-preven…] indexed:0 read:5min 2026-11-18 · —