The real threat from AI, as I see it, is not job replacement, or that it will take over the world. These threats get headlines and therefore woo investors. I think the real threat is already happening and it’s causing damage.
I am a fan of AI. Not a mad superfan, but I use it daily for things like troubleshooting, analysis, and spellchecking, including for this blog. It’s remarkable how well it can generate code and software generally.
As someone who builds and maintains websites, I find it invaluable that you can upload an error log and let it chew through it in seconds, rather than hunting for timestamps and error codes and trawling through forums and documentation for the solution. It’s now a case of upload, scan, done. The fix usually still requires technical know-how, but the solution is so much quicker to find that it is saving hours and maybe days of work every month. As a fan of productivity, that is an absolute game-changer.
But the biggest threat lies in how quickly it can produce code: it puts the power to build in the hands of people who have little idea what they are doing. I admit to having played around building chatbots, assistants and tools that are useful to me. This is how The Muster started. But I’m coming at this with a ten-year background in web development. I understand the server setups and the security protocols, but more importantly I know what I don’t know.
Recently someone told me AI had built them a website for free. They used a major AI app-building platform to build what looked like a professional website, with all the features they needed and automations to simplify operations.
I build WordPress websites, and I don’t believe this is a threat to that, because most of the people who would use these tools would have been using Wix or another easy builder. [LINK: /ai-threat-is-a-myth/ : same position as that post, so link it, then delete this note]
The site owner asked me to create an account on the site, and that was when I found the login flow invited me to use a different email address: entering an email address that didn’t match gave an error telling you which email to use, and from there the password could be set. This is a website with data and payments on it. Obviously I notified the site owner immediately, and within a few hours it was fixed. But the issue is that whoever built it with plain-English prompts did not know to check for these things.
Given the complexity of the site, it is likely other security features have been missed, such as an API key in a public folder, account files in a public folder, or something else.
I started looking into this. With platforms like Retool, Softr, Lovable and Zite, to name just four, anyone can create a free account and start building with no prior knowledge, then put the apps into production.
Who is checking these builds? As far as I can tell, no one. They are built and shipped with no technical oversight.
In May 2026, Israeli security firm RedAccess told Axios it found 380,000 publicly accessible assets built with tools from Lovable, Base44, Replit and Netlify, about 5,000 of them containing sensitive corporate data. Researchers said some of the tools default new apps to public unless users change it. Replit and Wix, which owns Base44, dispute parts of this, but the point stands.
The real risk with AI tools is not the tools: it’s the lack of oversight and the willingness of people who don’t know better to create tools that expose other people’s data.
So, some advice to anyone vibe coding, or hiring someone to build something for them with AI tools: what prior knowledge does the person building it have? Can they code? Have they built digital products and tools before? If not, you should probably find someone who has, or get someone to check it before you put it live.