6 min read
When did you last check what your AI add-ons can actually do on your computer? Most solo owners have installed at least one: a browser extension that drafts emails, a plugin for a writing app, or a “connector” that lets a chatbot read your files. We click “Add” in seconds, and we rarely think about it again.
A recent incident shows why that habit deserves a second look. In its November 13 security update, the AI coding company Cursor explained that on October 29 an attacker placed tampered copies of Anthropic’s Claude Code extension (versions 2.1.312 and 2.1.313) in the cache that serves Cursor’s extension marketplace. The tampered copies were available for about 11 hours, and up to 5,000 users may have received them. Anthropic did not publish those builds, and the flaw was not in Anthropic’s product. The attacker got in through an access key that a third-party vendor had exposed. As BleepingComputer reported, the extension looked legitimate because it carried the right name.
You may never open a coding tool. The lesson still applies to you: an add-on that looks official can still be dangerous, and the safest add-on is one you checked before trusting. Here are eight quick wins you can finish this week, most of them in under ten minutes.
Eight fast habits for safer AI add-ons
1. Make a list of every AI add-on you use
Open your browser’s extensions page, your writing apps’ plugin menus, and the “connectors” or “apps” section of each AI chatbot you use. Write each one down in a simple note: name, what it does, and when you last used it. You cannot protect what you have forgotten you installed. Most people find at least one they no longer recognize.
2. Remove anything you have not used in 60 days
Every add-on is a door into your browser or your files. If you have not opened it in two months, uninstall it. You can always reinstall it later. This single step shrinks your risk more than any setting.
3. Read the permissions, in plain words
Before you install, the browser shows what an extension can access. “Read and change all your data on all websites” means exactly that: it can see your bank, your email, and your client portals. Ask yourself whether the job it does needs that much access. A grammar helper might. A tab organizer probably should not. Our ChatGPT browser extension review walks through which permissions to tighten for everyday admin work.
4. Install only from the official source
Download extensions from the official store for your browser, and connectors from inside the AI app itself, never from a link in an email, a forum post, or a “free premium version” site. The Cursor case shows that even official stores can be poisoned, but they are still far safer than random downloads, and they remove bad listings fast.
5. Check the publisher, not just the name
Copycat add-ons often use the same name and logo as the real one. Before installing, look at who published it, how many users it has, and when it was last updated. A famous name with only a few hundred installs is a red flag. When in doubt, follow the link from the company’s own website to its store listing.
6. Turn on update alerts, then actually read them
Most extensions update silently. That is usually good, because updates fix bugs. But it also means a compromised update can arrive without you noticing. Once a month, skim your extensions page for anything that changed its permissions. If an add-on suddenly asks for more access than before, pause and find out why before accepting.
7. Keep client work in a separate browser profile
Create a second browser profile for client portals, banking, and anything sensitive, and install no AI add-ons there. Use your everyday profile for research and drafting with AI tools. It takes five minutes to set up and keeps your most important logins away from extensions that do not need them. This pairs well with the advice in our beginner’s guide to prompt injection, the trick where hidden text on a web page tries to give your AI new instructions.
8. Know your “if it goes wrong” steps in advance
Cursor’s advice to affected users included a line every solo owner should remember: “Uninstalling the extension alone may not be enough.” If an add-on you use is ever reported as compromised, the steps are the same:
- Uninstall it and restart your browser or app.
- Change the passwords for accounts you used while it was installed, starting with email and banking.
- Revoke and replace any API keys (the long secret codes that connect one service to another) that the tool could see.
- Follow the vendor’s official notice, which may include extra cleanup steps.
Write these four steps at the bottom of the list you made in tip one, so you are not working them out in a panic.
Why this matters more for a one-person business
A large company has an IT team that approves extensions, watches for strange activity, and cleans up after an incident. You are the IT team. That sounds daunting, but it also means you can move fast: a solo owner can audit every add-on in a single afternoon, something a 500-person company cannot do.
AI tools are also asking for more access than older software did. Connectors that read your inbox, your files, or your CRM are genuinely useful, and they are exactly the kind of access an attacker wants. That is not a reason to avoid them. It is a reason to connect them on purpose, one at a time, and to know which ones you have running.
If you want to turn these habits into a short written rule for yourself (and for any contractor who touches your accounts), our one-page AI policy guide includes a template line for approved tools and add-ons.
Your 30-minute add-on checkup
If you only have half an hour this week, do this:
- List every AI add-on and connector (10 minutes).
- Remove the ones you have not used in 60 days (5 minutes).
- Check the publisher and permissions of the ones you keep (10 minutes).
- Write your four “if it goes wrong” steps at the bottom of the list (5 minutes).
That is it. You do not need to be technical, and you do not need to give up the tools that save you time. You just need to know what is running, who made it, and what it can touch.
How many AI add-ons did you find when you made your list? Share your number in the comments; we suspect most readers will be surprised.



