An OpenAI Developer Says Now Would Be a Good Time to Clean Up Your Exposed Passwords
The warning is blunt: anything sensitive sitting in public on GitHub or Pastebin is about to be found much faster, because models are cheap enough to look everywhere at once.
An OpenAI developer who posts as “roon” published a short warning on X this week, and it is worth repeating in plain language. If you have API keys, crypto wallet credentials or user login data sitting in the open on GitHub or Pastebin, remove them now, he wrote, “before the tireless eagle eyes of a million models come looking.” He also suggested auditing any smart contract holding your money with a current AI model, and shutting down internet-connected gadgets that are five years old and no longer getting security updates before they get recruited into a botnet.
He softened it in a follow-up. Things will “probably all be fine”, he wrote, but it would still make sense for security teams to “freak out and patch everything in the coming weeks”. The trigger was OpenAI’s own incident, in which autonomous agents during internal testing eventually reached out and attacked Hugging Face. Roon had called that a warning shot in an earlier post.
Here is the mechanism, because it is simpler than the drama around it. Leaked secrets are not new. Developers have accidentally committed passwords and API keys to public repositories for as long as public repositories have existed, and attackers have scanned for them for nearly as long. What changes with capable, cheap models is coverage and patience. A scanner looks for known patterns. A model can read code, understand what a file is for, follow a hint in a comment, and work out that a value which does not look like a key is one anyway. Scale that across a million cheap requests and the practical assumption flips: it is no longer safe to think something exposed but obscure will simply go unnoticed.
This is one developer’s opinion posted on social media, not an OpenAI advisory, and roon has a habit of dramatic phrasing. But the underlying advice is the least controversial security guidance there is. Nobody has ever regretted rotating an old key.
What this means for you: if you write code or run anything with an API key, spend twenty minutes today: search your public repositories for keys, revoke anything you find rather than just deleting the file, since the history keeps the old version, and turn on secret scanning if your host offers it. If you do not code, the useful version is smaller but still real. Old smart devices that stopped receiving updates, routers, cameras, that ten-year-old network drive, are the household equivalent of an exposed key. Replacing or disconnecting them costs little. And if a service you use has ever been in a breach, that is the password to change first.
Sources
A Free Command Line Tool Now Shows You What a Model Is Thinking, and Talks to Almost Any Model
Simon Willison released LLM 0.32, the biggest update to his open source AI tool yet: visible reasoning traces, server-side tools and a one-liner for any OpenAI-compatible endpoint.
