Cursor 0-Day Disclosed, Vercel Locks Down Deployments

A Cursor zero-day went public, Vercel shipped deployment policies and encrypted SNI, and Lovable added Microsoft sign-in. What it means for your AI-built app.

Barret4 min read

Three things moved today that matter if you shipped an app with an AI builder. A researcher went public with a Cursor zero-day after quieter channels didn't work. Vercel added a way to lock down who can trigger a deployment, and turned on a privacy feature that hides which hostname a visitor is connecting to. Lovable, meanwhile, added another sign-in option to its auth settings. None of this is dramatic on its own. Together, it's a reminder that the attack surface of an AI-built app is bigger than the code you can see.

TL;DR

  • A security researcher published full details of a Cursor zero-day after disclosure channels reportedly failed (mindgard.ai)
  • Vercel shipped Deployment Policies, letting teams restrict which sources, orgs, and repos can push to production (Vercel changelog)
  • Vercel CDN now supports Encrypted Client Hello (ECH), hiding the hostname a client connects to from network observers (Vercel changelog)
  • Lovable added Microsoft sign-in alongside Google and Apple, widening the auth surface founders need to configure correctly (Lovable changelog)

The Cursor 0-day nobody could report quietly

A post making the rounds on Hacker News, from security firm Mindgard, argues that full disclosure became the only option left for a Cursor vulnerability (mindgard.ai). The post drew over 450 points and 200 comments, which tells you the developer community is paying attention. We don't have the technical details of the flaw beyond what's in that post, and we won't guess at them here. If you build with Cursor, the practical move is simple: read the disclosure yourself, check whether Cursor has shipped a fix, and update. Don't wait for a patch note to explain the risk to you after the fact.

Vercel gives you a way to restrict who can deploy your app

Vercel's new Deployment Policies let teams control which mechanisms, organizations, and repositories are allowed to trigger a deployment, configurable per environment at the team and project level (Vercel changelog). This matters more than it sounds. In a vibe-coded workflow, deployment often happens through a chain of integrations: a coding agent pushes to a repo, a webhook fires, a build runs, and code goes live with no human in the loop. Every link in that chain is a place someone could push code you didn't write. If you're running agents that commit and deploy on your behalf, this is the control that lets you say exactly which sources are trusted.

Less metadata leaks in the TLS handshake

Vercel CDN also turned on Encrypted Client Hello, which encrypts the Server Name Indication field in the TLS handshake — the part of an HTTPS connection that used to reveal which hostname a visitor was connecting to (Vercel changelog). With ECH on, network observers see a connection to Vercel's shared ECH hostname instead of your app's domain. It's automatic where supported and doesn't require any change on your end. It's not a fix for a vulnerability — it's a privacy improvement to how connections look on the wire, and it applies at the platform level for domains on Vercel DNS.

Lovable widens the front door

Lovable added Microsoft sign-in as a third option next to Google and Apple, managed through Cloud → Users → Auth settings (Lovable changelog). Lovable manages the OAuth credentials by default, so there's no Azure setup required unless you switch to your own credentials. Adding a sign-in provider isn't a vulnerability. But every provider is a new configuration to get right — redirect URIs, scopes, and what happens to a user record if someone signs in with a different provider using the same email. If you turn this on, check your auth settings the same day, not after a support ticket tells you something's off.

FAQ

What is the Cursor 0-day full disclosure about?

A researcher at Mindgard published details of a Cursor vulnerability publicly, arguing that other disclosure paths hadn't worked. The specifics of the flaw are in their post, not summarized here — read it directly and check for a Cursor update before assuming you're covered.

Should I restrict who can deploy your app on Vercel?

If a coding agent, CI pipeline, or external contributor can push to your repo, yes. Vercel's Deployment Policies let you limit which sources, orgs, and repos can trigger a production deployment, which closes a gap that's easy to overlook in an automated build chain.

Is Lovable's Microsoft sign-in secure by default?

Lovable manages the OAuth credentials for you unless you switch to your own Azure app, so the default setup is reasonable. The risk isn't the provider — it's misconfiguration. Check your Auth settings after enabling it, and confirm how accounts are matched across providers.

The bottom line

Today's changes aren't a breach story. They're a reminder that the tools you build with keep adding surface area — new auth providers, new deployment paths, new ways your traffic can be observed or intercepted. None of that is bad. It's just more for you to check. A zero-day disclosure, a new deployment control, and a new sign-in button all point to the same habit: read the changelog, and don't assume a platform default is a security review you already did.

Find your gaps before an attacker does.

Is My Site Hackable? scans your deployed app for the exact issues in this article — exposed keys, missing RLS, open buckets — and tells you what's real and what's a false alarm.

Run a free scan →