StackAttestTechnical Trust
AI builders

Is Bolt.new safe? What the platform secures and what you must

Updated November 15, 20268 min read
Short answer

Bolt.new is a legitimate platform with real security controls: Bolt says it is SOC 2 Type 2 certified, and it can review each app before it goes live. Whether your app is safe is a separate question. The gaps that matter sit in what Bolt generated: where permission checks run, which keys reach the browser, how webhooks are verified, and the database rules behind the app.

Mapped toOWASP ASVS 5OWASP API Security Top 10CWE

Two questions hide inside "is Bolt.new safe". One is whether the platform can be trusted with your code and your account. The other is whether the application it built for you is safe to put in front of strangers. The first has a short answer. The second is where launches go wrong.

Is the Bolt platform itself safe?

On its security page, Bolt says it is SOC 2 Type 2 certified and compliant with GDPR and CCPA. It publishes a trust profile at trust.bolt.new with audit reports and policies, and a way to request documentation for a security review.

Bolt can also review every app before it goes live and apply a fix for issues it finds, once you switch that on. Switch it on. It is a useful first filter and Bolt does not charge extra for it.

None of that makes the generated application safe by itself. A SOC 2 report describes how Bolt runs its own systems. It says nothing about the permission checks inside your app, because those were written for your project, in one pass, and only you know what they should allow.

Where Bolt apps usually need work

These are failure modes common to any app generated in one pass, not claims about yours. Each one can be checked in minutes, and together they cover what attackers try first.

1. Permission checks that live in the interface

A generated app tends to enforce rules where they are visible: the admin button only renders for admins, the edit form only appears for the owner. A request sent straight to the endpoint never meets those checks. Sign in as one user, copy a request that names another user's record, and replay it. If the record comes back, ownership is not enforced on the server.

2. Keys that reached the browser

Bolt apps are usually built with Vite or Next.js, where a variable named with the VITE_ or NEXT_PUBLIC_ prefix is copied into the code every visitor downloads. That is fine for a publishable key and fatal for a secret one. Open the deployed site, search the loaded sources for secret, service_role and your providers' names, and rotate anything privileged you find there.

3. The database rules, if Supabase sits behind it

Many Bolt apps store their data in Supabase. A Supabase table without row level security is readable by anyone holding the public key, and that key is in the browser by design. List your tables, confirm row level security is on for each, and check that every policy compares the row's owner to the signed-in user rather than just checking that someone is signed in.

4. Webhooks wired until the demo worked

A payment webhook that acts on whatever it receives lets anyone who finds the address announce a payment that never happened. Post a plausible payload with no signature and confirm it is rejected, then send the same signed event twice and confirm it only took effect once.

5. Dependencies that arrived in one step

A whole stack of packages appears in a single generation, none chosen individually. Run your package manager's audit command and treat critical and high advisories as launch blockers. Most close with a version bump.

Is Bolt.new safe for production?

The platform is a reasonable base for production. Your app is ready when the five checks above pass and someone has tried to read another account's data with a second login and failed. If the app takes payments, stores health or financial data, or is about to face a customer's security review, get an independent check before launch rather than after.

Test your deployed Bolt app from outside, free and without an account.Run the free check

When you need evidence someone else will accept, StackAttest tests these controls against named standards, from the deployed app, the source and the running system, and records how strongly each result was established.

See what an independent audit of a Bolt app covers.Bolt security audit

Frequently asked questions

Is Bolt.new SOC 2 certified?

Bolt says it is SOC 2 Type 2 certified and compliant with GDPR and CCPA, and it offers audit reports through its trust profile at trust.bolt.new. That certification covers Bolt's own systems, not the apps you build with it.

Is Bolt.new safe for medical or sensitive data?

Bolt's security page mentions HIPAA only for running Bolt inside your own cloud tenant, not for its standard cloud service. If you handle health data, confirm the position with Bolt directly, and have the app itself reviewed before it holds real records.

Does Bolt's built-in review replace an audit?

No. It is a good first filter and worth switching on. It reviews what Bolt generated. It does not sign in to your live app as a second user, and it does not produce evidence someone outside your team can check.

Can Bolt apps be hacked?

Any public app can be attacked. Bolt apps are not unusually weak, but they inherit decisions nobody reviewed, and the five checks above close the gaps attackers try first.

Keep reading

Is Lovable safe? The platform, and the app you built on it

Guide · 10 min

Is Supabase safe for production? Yes, once you do your half

Guide · 9 min

Who can check your AI-built app before launch, and how

Guide · 12 min

The vibe coding security checklist: 12 checks before you ship

Article · 9 min

Why AI-generated code ships with security flaws, and what to do about it

Article · 6 min