Back to Insights

Researcher Reports 35,000 Exposed Grok Build Apps: What Businesses Using AI Coding Tools Should Check

4 min read

What Happened

According to a report from Ars Technica, a security researcher found roughly 35,000 apps built with Grok Build, xAI’s AI coding agent, sitting wide open to anyone who went looking for them. The report says those exposed apps leaked a significant amount of confidential information, and that some of it belonged to SpaceXSI.

The details here come from the researcher’s findings as reported by Ars Technica. We haven’t seen independent confirmation of every specific, and we’re not going to guess at exactly how each app was exposed or what each one contained. But the headline finding is clear enough to be worth your attention: tens of thousands of apps built quickly with an AI coding tool were reportedly accessible to people who were never supposed to see them.

Why Business Owners Should Care

You might be thinking, “I don’t use Grok Build, so this isn’t my problem.” That’s partly true. But the bigger lesson isn’t about one product. It’s about a pattern.

AI coding agents have made it remarkably easy for anyone on your team to spin up a working app in an afternoon. A sales manager can build a quote calculator. An office administrator can create an internal dashboard that pulls in customer records. That’s genuinely useful, and it’s a big part of why these tools are taking off.

The catch is that “it works” and “it’s safe” are two different things. An app that runs perfectly can still be published to the open internet, store data without proper access controls, or have passwords and keys baked right into it. The people building these apps often aren’t developers, so they may not know which questions to ask. And because the tools are so fast, apps can multiply across a business before anyone in charge knows they exist.

When you hear that a single researcher reportedly found 35,000 exposed apps from one tool, the takeaway isn’t that this tool is uniquely bad. It’s that exposure at scale becomes possible when building gets this easy and checking doesn’t keep up. If a well-resourced organization’s information reportedly ended up in that pile, a small business with no dedicated IT security person is not immune.

We’ve written before about how AI systems can act in ways their owners didn’t intend in our look at rogue AI agent activity and what businesses should check. This story is a cousin of that one: the risk isn’t always the AI misbehaving. Sometimes it’s people moving faster than their safeguards.

Practical Takeaways

1. Find out what’s already been built

Before you worry about policy, take inventory. Ask your team a simple, non-judgmental question: have you built or published anything with an AI coding tool, for work or using work data? You may be surprised. You can’t protect apps you don’t know about, and people are more likely to be honest if they don’t feel they’re in trouble.

2. Check who can actually reach each app

For every app you find, answer three questions. Is it on the public internet or only accessible internally? Does it require a login? Does it touch customer, financial, or employee data? Anything that’s public, has no login, and handles sensitive information should be taken down or locked down until someone qualified reviews it.

3. Set a simple rule before the next app gets built

You don’t need a 40-page policy. A short rule works: AI-built tools that handle real business data need a second set of eyes before they go live. That review can be a trusted IT provider, a technically minded employee, or an outside consultant. The point is that nobody publishes something with customer data on their own judgment alone.

What to Watch

Keep an eye on how xAI responds to the report and whether it changes default settings in Grok Build, such as making apps private unless a user deliberately publishes them. More broadly, watch whether other AI coding platforms tighten their defaults. Safer defaults from the vendors would help, but they won’t replace knowing what your own team is building.

The PAD Take

This week, spend 30 minutes asking your team what they’ve built with AI coding tools and listing every app that touches customer or company data. Lock down or pull anything public that handles sensitive information, and require a review before anything new goes live. If you’d rather have someone walk through your current AI tools and flag the gaps, our AI checkup for reviewing the tools your business already uses is built for exactly this.


More from the blog

Enjoyed this? Get more like it.

Practical AI insights, delivered monthly. No spam.

Unsubscribe anytime. We respect your inbox.

All Insights