Fundamentos de segurança para código de IA
Evite os riscos que a IA frequentemente não identifica.
Fundamentos de segurança para código de IA é uma aula grátis de Vibe Coding no CoddyKit. Esta é a aula 2 de 4. Você pode ler a aula completa abaixo gratuitamente — depois pratica ao vivo no navegador com um editor de código integrado e um tutor de IA 24/7. Faz parte do caminho de aprendizado de Vibe Coding, e seu progresso é sincronizado entre a web e o app CoddyKit. O curso de Vibe Coding inclui 4 aulas no total.
Partes desta aula ainda não foram traduzidas e aparecem em inglês.
AI Writes Fast — Sometimes Unsafely
AI is brilliant at getting something working quickly. But "working" and "safe" aren't the same. AI often leaves real apps open to attacks because it optimizes for the happy path, not for hostile users.
The good news: a handful of habits catch most problems, and you can ask AI to apply them. This lesson covers the security risks vibe coders hit most.
Never Put Secrets in Code
The #1 mistake: pasting an API key or password directly into your code. If that code goes to GitHub, the key is public forever — bots scan for leaked keys within minutes.
Secrets belong in environment variables, never in the source. Spot the difference below.
// DANGER: key is hardcoded and will leak
const apiKey = 'sk-live-abc123realkey';
// SAFE: key comes from the environment, not the code
const apiKey = process.env.API_KEY;
if (!apiKey) {
throw new Error('API_KEY is missing — set it in your .env file');
}Keep Secrets Out of Git
Store secrets in a .env file, and make sure Git ignores that file so it never gets committed. One line in .gitignore does it.
Ask your AI tool to set this up correctly from the start.
Set up environment variables for this project:
- create a .env file with API_KEY=
- add .env to .gitignore so it's never committed
- create a .env.example with empty placeholders to commit instead
- show me where the code reads process.env.API_KEYNever Trust User Input
The golden rule of security: anything a user can type, a user can lie about. Attackers send broken, oversized, or malicious input on purpose. Your code must validate it.
Here's a basic guard that checks input before using it.
function createUser(input) {
if (typeof input.email !== 'string' || !input.email.includes('@')) {
return { error: 'Invalid email' };
}
if (input.age < 0 || input.age > 150) {
return { error: 'Invalid age' };
}
return { ok: true };
}
console.log(createUser({ email: 'bad', age: 30 }));
console.log(createUser({ email: 'a@b.com', age: 30 }));SQL Injection: A Classic AI Slip
If AI builds your database query by gluing user input into a string, an attacker can inject their own commands and steal or delete data. This is SQL injection.
The fix is parameterized queries — the database treats input as data, never as commands. Always ask AI to use them.
// DANGER: user input glued straight into the query
db.query("SELECT * FROM users WHERE email = '" + email + "'");
// SAFE: parameterized — the ? is filled in safely
db.query('SELECT * FROM users WHERE email = ?', [email]);Ask AI to Review for Security
You can turn the AI into a security reviewer. Paste your code and ask it to hunt for vulnerabilities specifically — it's much better at finding them when you ask directly.
Review this code for security problems. Specifically check for:
- hardcoded secrets or API keys
- unvalidated user input
- SQL injection
- missing authentication checks
List each issue, why it's risky, and the fix. Here is the code: <paste>Protect What Should Be Private
AI often builds an endpoint that returns data without checking who is asking. That means anyone could read another user's data just by changing an ID in the URL.
Every sensitive action needs an authorization check: is this user allowed to do this?
// DANGER: anyone can read any order by guessing the id
app.get('/orders/:id', (req, res) => {
res.json(getOrder(req.params.id));
});
// SAFE: only the owner can see their order
app.get('/orders/:id', (req, res) => {
const order = getOrder(req.params.id);
if (order.userId !== req.user.id) return res.status(403).end();
res.json(order);
});Keep Dependencies Updated
Your app uses many packages written by others. Some have known security holes. Outdated dependencies are one of the most common ways apps get hacked.
A quick command checks for known vulnerabilities, and AI can help you understand the results.
# Check installed packages for known vulnerabilities
npm audit
# Then ask AI:
# "npm audit reports 3 high-severity issues in <package>.
# Explain the risk in plain English and tell me the safest way to fix it."Don't Leak Errors to Users
When something breaks, AI-generated code sometimes shows the raw error — including file paths, database details, or stack traces. That's a gift to attackers.
Log details privately on the server, but show users a generic, friendly message.
try {
doRiskyThing();
} catch (err) {
console.error('Internal error:', err); // private log
res.status(500).json({ error: 'Something went wrong. Please try again.' });
// never send err.stack or err.message to the user
}A Security Checklist You Can Reuse
Before you ship anything, run through this. Paste it to your AI and ask it to verify each item against your code.
- No secrets in the code or Git history.
- All user input validated.
- Database queries are parameterized.
- Sensitive routes check who is logged in and what they're allowed to do.
- Dependencies audited and updated.
- Errors logged privately, generic messages to users.
Security Is a Conversation, Not a One-Time Step
You won't catch everything, and that's fine. The mindset that protects you: assume AI code is insecure until reviewed, ask security questions every time you add a feature, and re-run your checklist before each deploy.
Treat the AI as a careful partner you actively interrogate — not an oracle you blindly trust.
Quick Check
Your AI assistant generates a login feature and hardcodes the database password right in the source file. What should you do?
Recap: Safe by Habit
AI ships fast but skips safety, so you supply the safety. You learned to: keep secrets out of code, validate all input, use parameterized queries, add authorization checks, audit dependencies, and hide raw errors.
Best of all, you can ask AI to review for these exact issues and run a checklist before every deploy. Next: keeping your code clean and maintainable.
Perguntas Frequentes
A aula “Fundamentos de segurança para código de IA” é grátis?
Sim — o texto completo de “Fundamentos de segurança para código de IA” é grátis para ler aqui na web. Para praticá-la interativamente (um editor de código integrado e um tutor de IA 24/7) e desbloquear o restante do curso de Vibe Coding, atualize para CoddyKit PRO. O curso de Vibe Coding inclui 4 aulas no total.
O que vou aprender em “Fundamentos de segurança para código de IA”?
Evite os riscos que a IA frequentemente não identifica. Você pratica Vibe Coding com código prático que executa diretamente no navegador, e um tutor de IA 24/7 responde suas dúvidas enquanto trabalha na aula.
Preciso ter experiência prévia para começar Vibe Coding?
Nenhuma experiência prévia é necessária. Vibe Coding no CoddyKit é estruturado para alunos iniciantes até avançados, então você pode começar aqui ou desde o início e aprender no seu ritmo. Esta é a aula 2 de 4.
Quanto tempo leva a aula “Fundamentos de segurança para código de IA”?
A maioria das aulas CoddyKit leva cerca de 5–10 minutos. Cada uma é compacta e interativa, então você faz progresso constante e retoma exatamente de onde parou entre web e app.
Posso escrever e executar código nesta aula de Vibe Coding?
Sim. Cada aula de Vibe Coding inclui um editor de código integrado, então você escreve e executa código real direto no navegador e recebe feedback de IA instantaneamente — nenhuma configuração local necessária.
Todas as aulas deste curso
- Adicionando testes com IA
- Fundamentos de segurança para código de IA
- Mantendo o código sustentável
- Desempenho e custo