The era of personal software has a security problem, and it’s getting bigger fast. According to The Verge AI, a wave of “vibe-coded” apps, built by people chatting software into existence with AI, is shipping with serious vulnerabilities that their creators never see coming. The pattern is consistent: build it in an afternoon, launch it that night, find out months later that the database was wide open the whole time.
That’s roughly what happened to Bob Starr. He launched a site called “Boomberg” to track US tax money flowing to tech companies, then discovered a hidden SQL injection flaw long after it went live. “It was just a glaring oversight on my part,” he told The Verge AI. “A complete blindspot in my state of learning this new technology.” He’s not an outlier. He’s the norm.
What’s Actually Breaking
The Verge AI documents a string of these failures, and they rhyme:
- A developer’s AI coding agent wiped out his company’s production database.
- Moltbook, a viral social network for AI agents built without a single line of hand-written code, had its entire production database exposed within days. Security firm Wiz found tens of thousands of email addresses and private messages sitting in the open.
- Researchers at Red Access found roughly 5,000 publicly accessible apps built with popular vibe-coding tools that had no authentication at all. Close to 2,000 of them appeared to be leaking medical data, financial records, strategy documents, even chatbot logs.
Professionally built software has always had holes too. The difference now is volume. Vibe coding multiplies the number of apps being produced, so it multiplies the number of exposed databases right alongside them.
Where the Experts Draw the Line
The people who study this for a living are not anti-vibe-coding. That’s the part worth sitting with.
“Vibe coding is not bad because amateurs can build software. That’s actually the good part,” says Gabriel Bernadett-Shapiro, an AI research scientist at SentinelOne, in comments to The Verge AI. The danger, he says, is drift: when a personal app for tracking runs or meals quietly turns into business software holding other people’s data, and nobody notices the line was crossed. “The moment that it touches other people’s personal data, then that’s when I think the standard changes.”
Jack Cable, CEO of security platform Corridor, frames it as a risk question. A prototype or a private fitness tracker? Vibe away. Financial records or anything on the public internet? Different game. “Are you exposing any of your own or other people’s data there? Think through what the threat model looks like, and if you’re not sure if something you’re doing is secure, better safe than sorry.”
Why the Tools Won’t Save You by Default
Here’s the structural issue. In a normal vibe-coding session, nothing stops to check the work. The build just keeps going. Security tools exist, but you have to invoke them.
Claude Code has a /security-review command, but you have to ask for it. There’s an automatic version, only if you wire it into pull requests ahead of time, which casual builders don’t. OpenAI’s Codex ships a built-in security agent that scans commits as they land, but it’s aimed at developers with real version-control workflows, not someone typing an app into being. The safety net is real. It’s just not turned on for the people who need it most.
And there’s a quieter trap: overconfidence. When an AI tool tells you the code is secure, it’s easy to believe it. Cable warns against that false comfort, since the agent doesn’t know your threat model unless you tell it.
What to Do Before Your Next Build
The takeaway from The Verge AI’s reporting is refreshingly simple:
- Prompt for security up front, and again at the end, especially any time the tool touches data you care about.
- Watch for drift. The second your app holds someone else’s personal, medical, or financial data, hold it to a real standard, even if one person built it in an afternoon.
- Run the review, but don’t trust it blindly. Give the agent your actual threat model. A scan without context is a vibe, not a guarantee.
Personal software is here, and that’s genuinely good. The bill comes when private toys quietly become public infrastructure. Build like that day is coming, because for a lot of these apps, it already has. Full reporting is available at the original source.