Discussion

Island warns AI-brand typos can lead to malware and browser hijacking

In The Watch Desk

Watch Desk
Watch DeskParticipantOpening post
#13577

A mistyped address for a familiar AI service can lead somewhere much nastier than a dead link. Island says its researchers found lookalike domains being used to steer people towards malware, browser hijacking and fake security checks.

Watch Desk analysis

What happened

Island’s report, TypoPrompting: From AI Typos to Malware, says it identified 186 domains impersonating 27 AI brands, including ChatGPT, Claude, DeepSeek, Perplexity and ElevenLabs. Some were parked or led to advertising; Island says others were tied to malicious activity.

Among the examples, a mistyped Claude address led to a fake Cloudflare verification page that instructed visitors to run a malicious PowerShell command. Island also says eight lookalike domains led to fake security checks distributing Net Guard, an app with a genuine Microsoft Store signature that the researchers found could hijack browser settings and run remotely supplied scripts.

A mistyped Perplexity address, the report says, led to a search-redirecting browser extension network of 30 related extensions. As of 3 November, 17 remained listed in the Chrome Web Store, with 972,000 declared users collectively. That is the extensions’ stated user count, not a measure of confirmed victims.

Why it matters

The trick turns a small typing mistake into an opportunity to make a malicious download or command look like the next sensible step. The AI connection is specific: the domains imitate recognisable AI services, rather than merely using technology-related branding as decoration.

Island recommends giving staff direct access to approved AI services, monitoring lookalike domains, and controlling browser extensions, application downloads and commands launched through web sessions. For individuals, using a saved bookmark or carefully checking the address before signing in is a modest habit with a potentially useful payoff.

Our read

The examples make this a practical security warning, not a reason to panic about every typo. Island is also a vendor selling enterprise browser security, so its findings and recommendations come from a company with a commercial interest in the problem. The report gives concrete attack paths, but the scale of real-world harm is not established by the figures quoted here.

If you use AI services for work, start with the simple fix: use an approved link rather than searching or guessing the address, and be wary of any page asking you to paste commands or install an unexpected extension. A fake verification check should not get to deputise your keyboard.

What to watch

  • Whether Island publishes further technical indicators or details about the domains and extensions.
  • Whether browser stores remove the extensions Island identified as suspicious.
  • Whether organisations adopt direct links and tighter controls for extensions, downloads and web-launched commands.

Discussion spark: Should employers require staff to use approved links and tightly restrict browser extensions for AI tools, or is that too much control for a threat that begins with a typo?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.