{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreif42rtsgfnxltmfbo72h4xs4gr2qbqlvv2khrvoxpzkoko2br7oem",
    "uri": "at://did:plc:yamgjp3fq22zggtm5g7ply4n/app.bsky.feed.post/3mmpdxt66u3y2"
  },
  "description": "There's a growing trend of users inserting AI into production servers, but why? ",
  "path": "/signals/the-need-to-insert-ai-into-everything/",
  "publishedAt": "2026-05-25T20:47:09.000Z",
  "site": "https://artofinfra.com",
  "tags": [
    "numerous stories",
    "@levelsio",
    "@Tailscale",
    "@Hetzner_Online",
    "@claudeai",
    "@Cloudflare",
    "@TermiusHQ",
    "pic.twitter.com/Py6VuYYHw0",
    "May 24, 2026",
    "@DavidFrosdick"
  ],
  "textContent": "It seems to be commonplace right now that people who don't know what they're doing follow influencers into bad security habits. Right now, there seems to be a trend of inserting AI agents onto production servers. We all know how that ends.\n\nWe've seen the numerous stories of how AI agents have deleted databases; they've amended them in some certain way. Never mind the fact that these LLM agents are reading the data and processing it without consumers' or users' consent. Does that not concern you?\n\nTake this story, for example, on X or Twitter as best known. levelsio is sharing the fact that he utilises AI agents on his production servers, managing his databases, his production app, and having a follower base like he does, you are inherently leading people to bad practises.\n\n> I did it 👍 thanks @levelsio for sharing.\n>\n> Setup @Tailscale on my @Hetzner_Online VPS 🙌\n>\n> Installed @claudeai on VPS 🤞\n>\n> All my sites run through @Cloudflare tunnels 😎\n>\n> Got @TermiusHQ running locally on my Mac and iPhone ❤️\n>\n> Locked all the ports 💪\n>\n> Migrate 25GB of projects,… pic.twitter.com/Py6VuYYHw0\n>\n> — David Frosdick (@DavidFrosdick) May 24, 2026\n\nNot only can these AI agents read your API keys, but they can also malform your app in weird ways. They can read and access and destroy your database. But third, you don't actually know what they're doing behind the scenes. You don't.\n\nAs detailed above, you're probably not even disclosing to your users, or the people assigned up to your app that your data is now inherently in the hands of whatever company is processing the LLM agent's input, be it OpenAI, Anthropic, or anywhere in the world. These users do not know that their data is being utilised in this way.\n\nSecurity, my god security. What is security? In this day and age what is security? What does it mean? We know apps like Claude and ChatGPT allow remote access into agents if required and prompted for. So you are inherently opening the demon's doors, open access to your production servers. Does that sound like a good idea?\n\nPlease, if you are thinking about installing AI/ LLM agents, do not install them on your production servers. Worst-case scenario, if you REALLY need to, give them a user account (with READ-ONLY access), from a VM you manage locally or your laptop that you manage. Give them remote access and limit that user space in the operating system on your remote server. Just give them read access because these people right now are literally installing them with sudo access.\n\nDo not, I repeat, do not do that. Give them read-only access if you need to. I recommend not allowing them to touch your production servers whatsoever.\n\nDon't get me wrong, I'm a fan of AI myself. I use AI daily, but it comes a point where there's so much out there, so much false information, that anyone can be dragged into a situation that is dangerous, damaging, and also harmful to users.\n\nI'll be sharing soon a full post about how I interact with AI and best practices I've learned along the way. I've been using it now since probably the dawn of when ChatGPT was first announced. But I want to ensure that people who are here to learn from what I've learned in the last two or three years. So please don't follow the advice from influencers. Always question what you're doing, and I'll share more information soon.",
  "title": "Why do people need to insert AI into EVERYTHING?",
  "updatedAt": "2026-05-25T20:47:09.411Z"
}