26/10/2026
Building with AI? Here’s what to check before you go live. 🔍
Last week we looked at a site that was hacked six days after launch. The cause was a password left in plain sight. It’s not the only mistake AI tools make, so here are the ones we see most:
1. SQL injection
If user input goes straight into a database query, an attacker can type their own commands into a login or search box. They can then read, change or delete your data. The fix is parameterised queries, and AI-generated code doesn’t always use them.
2. Passwords and keys in the code
Admin passwords, API keys and database credentials hard-coded where anyone can read them. A browser’s developer tools is all it takes.
3. Pages that look protected but aren’t
Hiding the admin button is not security. If the server behind it doesn’t check who is asking, anyone can request the data directly, including other customers’ records.
4. Nobody who can explain the code
If no one on your team can say what the app does with customer data, no one is accountable for it. Under POPIA, that responsibility stays with you.
5. No testing before launch
“It works” and “it’s safe” are two different questions. Automated scanners start probing new sites within hours of going live.
AI is a great way to build faster. It just needs an experienced engineer reading what it produced.At {type}DEV, we can review your code before launch, or after the fact if you’re already live. 👇
📩 Send us your project and let’s have a look.
🌐 www.typedev.co.za
✉️ [email protected]