Top News

Artificial Intelligence (AI)’s habit of giving wrong answers, called ‘AI hallucination’, has become a new and serious cyber threat. Hackers are now taking advantage of this weakness of AI to cheat people through ‘AI squatting’. In this attack, hackers already guess which wrong website or software name the AI ​​can take. Then they buy those names themselves and when a user clicks on that link at the behest of AI, he directly gets trapped in the hackers’ trap.

AI tools sometimes create website links or software names that do not actually exist. Earlier people used to leave it as just a small mistake but now it has become a big security issue. Hackers are using these mistakes of AI as their weapon. They pre-register the names of websites and software about which AI is more likely to provide false information. When you trust AI and click on those links, you unknowingly reach a website from where your information can be stolen or your computer can get a virus.

Phantom Squatting Fake Trap

‘Phantom squatting’ means buying website addresses that AI creates by mistake. Unit 42 Research Team of Palo Alto Networks has explained about this in detail. Hackers first ask several AI questions to find out which fake websites they are calling. After this they check which website is currently vacant and buy it. Then they create a fake website on it which looks absolutely real. Once, people’s money and their important information were stolen by creating a fake website of a postal service. Even to create that fake website, the hackers had taken the help of AI coding assistant.

Danger in coding due to hellosquatting

The danger of ‘halloosquatting’ is there for people who do coding work. When a coding person asks AI to download software, AI often gives names of software that do not exist. Hackers put virus software on those fake names. When a developer runs the command given by AI in his computer, a virus gets infected in his computer. This is very dangerous because the possibility of this is very high in coding work. Research has also revealed that the cost of AI mistakes during coding can range from 85 percent to 100 percent. The hacker does not have to do anything in this, he just waits for the AI ​​to make a mistake.

Why is it difficult to prevent this mistake?

This problem cannot be completely fixed by any software update or patch, this is just the way AI works. AI does not check whether the website or software is genuine or not, it simply adds words and answers. This danger increases even more when AI agents start accessing the internet or downloading software automatically. Then these wrong answers of AI can become the way for a major cyber attack. There is no need for anyone to do any hacking in this, it is a part of the normal behavior of AI.

How to keep yourself safe?

This danger has now increased for common people as well as big companies. Research has revealed that hackers have created more than 13,000 dangerous links and more than 2,50,000 such domains in the name of 913 big brands which AI can identify by mistake. Now companies should check every link or software given by AI on the internet themselves before using it.