Seven million sites generated by AI on the Lovable platform expose users’ private data. Exposed data from AI-generated sites includes passport numbers and payment information. This is one of the biggest problems with the “democratization” of software engineering. Creating a website in minutes via conversation with AI is easy, but many forget about security. No wonder we’re seeing so many breaches. It’s a self-reinforcing cycle—the current level of AI is so advanced that it can uncover hidden bugs that have gone undetected for years. But the increased ease of generating apps also introduces new risks. Anyone creating an app or website with AI has a responsibility to conduct multiple adversarial reviews to try to “break” the app before launching it to the public. They also have a duty to continue maintaining it—something professionals understand, but those who merely “vibe-code” often overlook. Software is an ongoing process, which is why the rule “never ship on a Friday” exists—because you might end up spending your weekend buried in logs. Today, it’s possible to literally create an agent whose sole job is to monitor the app, perform QA, and open pull requests for any issues found. That’s a start. But it’s never acceptable to deploy and then forget that ongoing maintenance is required.