What The Tech: How cyber criminals are tricking artificial intelligence
BY JAMEY TUCKER, Consumer Technology Reporter
The hidden prompts can trick the AI assistant into revealing private information, downloading harmful code, or connecting to unsafe sites.
Say you search for ‘best gifts for a 9 year old’. If one of the pages that AI browses has hidden text, the AI might read it as an instruction, not just part of the page.
How is it hidden? The cyber criminals write the code in white text, on a white background. So it’s invisible unless you know to reverse the color.
Otherwise, you wouldn’t see anything wrong. But AI could follow those invisible commands. There are real examples of this happening in browsers like Chrome, Edge, or tools like Perplexity. Researchers have found they can be fooled into sending the sensitive data to hackers without the user’s knowledge.
Browser developers are racing to fix this but the technology is moving fast.
If you don’t want to use AI in your browser, you can turn it off. In Chrome, click the beaker icon in the upper right hand corner, open settings, and toggle off AI Mode and AI in Search.
These attacks are rare, but they are happening in the real world. An important reminder to keep browsers and the operating systems updated since they often include patches for new vulnerabilities. And what you can’t see, can cause a lot of problems.
