Consider the following: you instruct your AI browser to do some shopping on your behalf, but it's only able to purchase counterfeit goods on phony websites. That's exactly what occurred with Perplexity's Comet AI browser. Insane, right?
I've been curious about this for some time now—if I let AI do my shopping, will it be scammed? Well, we now know that the answer is yes.
The AI believed the imitation stores were genuine and went ahead with the purchase. It's as though you send a friend to the mall, but he gets himself into some dubious store that deals in fakes.
And let's be honest—we sometimes can't tell if a website is safe. If AI can't even perform any better than us, what's the point?
• It didn't examine the source quite as well.
• It doesn't have gut intuition like humans. We simply "feel" when something is off.
• It was too overconfident.
I've mindlessly clicked on busy "SALE" banner ads and nearly made a purchase, but my intuition intervened. Comet AI lacks that gut intuition.
• Always keep a human in the loop. AI must not spend money without approval.
• Improved safety measures. Comet must have improved means of checking whether a site is authentic or not.
• Be cautious. Use AI as an assistant and never as a dictator.
Because seriously—would you allow a robot to shop on your credit card if it can't tell a scam from the real thing? Nope, me neither.
I've been curious about this for some time now—if I let AI do my shopping, will it be scammed? Well, we now know that the answer is yes.
What Happened?
Comet AI is designed to simplify shopping. You tell it what you want, and it attempts to look for alternatives, even ordering for you. But scientists discovered that hackers deceived Comet to accept imitation sites.The AI believed the imitation stores were genuine and went ahead with the purchase. It's as though you send a friend to the mall, but he gets himself into some dubious store that deals in fakes.
Why Does It Matter?
You may think, "So what if it's fake shoes?" But here's the thing: if AI can't detect scams, it may fall for worse scams such as fake electronics, stolen credit cards, or phishing.And let's be honest—we sometimes can't tell if a website is safe. If AI can't even perform any better than us, what's the point?
What Went Wrong?
Here's why Comet failed:• It didn't examine the source quite as well.
• It doesn't have gut intuition like humans. We simply "feel" when something is off.
• It was too overconfident.
I've mindlessly clicked on busy "SALE" banner ads and nearly made a purchase, but my intuition intervened. Comet AI lacks that gut intuition.
What Can Be Done?
A few fixes can help:• Always keep a human in the loop. AI must not spend money without approval.
• Improved safety measures. Comet must have improved means of checking whether a site is authentic or not.
• Be cautious. Use AI as an assistant and never as a dictator.
If You Asked My Thoughts Then
I rather like the thought of my shopping being done by AI. I despise opening ten windows to hunt around for prices. But if it won't prevent me from being scammed, I'll continue to shop in my usual way.The Bottom Line
AI browsers are nice, but not quite clever enough yet to shop on their own. Until they can detect scams, we have to double-check.Because seriously—would you allow a robot to shop on your credit card if it can't tell a scam from the real thing? Nope, me neither.
