AI Agents Are Better Hackers Than Humans And It Is A Big Worry: Here’s The Full Story | Tech News

AI agents are working better as hackers and even cheaper. (Photo: AI generated)

AI agents are enabling people to multi-task which has become a pivotal part of the AI era but the technology has its perils and one of them is setting up a worrying trend. Security experts have warned about AI-powered attacks that can be damaging but now AI agents are disguising as hackers and apparently doing much better than humans at the task.

The details come via an experiment done by researchers at Stanford University and did so without having to spend big on recruiting human hackers. 

ARTEMIS AI Hacks Better Than Humans 

The AI agent named ARTEMIS has been developed to counter the threat of cyber hacking and show the world how the technology is getting better and that is scary. The researchers deployed the AI agent on the public and private computers running in the Stanford network for 16 hours and trained it to work across 8,000 devices to look for security issues and exploit them.

The results were mostly shocking because ARTEMIS proved itself to be better than skilled human hackers. The tests showed the AI agent doing better than 9 out of the 10 hackers it was competing with and even highlighted some of the weaknesses that the human counterparts couldn’t spot. 

The team led by Justin Lin, Eliot Jones and Donovan Jasper ran these tests over the Stanford network, most likely to prove a point about the power of AI agents and what they are capable of in the wrong hands. 

AI Agent Hackers Prove A Point 

The tests were run at a critical point wherein the AI agent was only trained to a certain level but it had the power to run for much longer than usual AI models. It managed to scan, probe for security issues in these devices and even deliver its analysis for a longer period. 

The accuracy of these findings during the test was over 80 percent which is impressive when you consider the AI agent was built at $18 (Rs 1,476 approx) per hour, while the human agents earned well over Rs 1.10 crore in a year. 

Google and Microsoft have deployed AI to scan and prevent cyber attacks and the ARTEMIS test shows you why. We can expect cybersecurity companies to adopt AI at a much deeper level but they should be aware and cautious about the possible mistakes these models will make during the process.

Click here to add News18 as your preferred news source on Google.