Unprecedented: Artificial intelligence penetrates its own kind.

 

The artificial intelligence company "Hugging Face" was hacked, but strangely enough, the perpetrator was an entirely artificial intelligence system

The artificial intelligence company "Hugging Face" was hacked, but strangely enough, the perpetrator was an entirely artificial intelligence system.

It is known that hackers are increasingly using artificial intelligence to discover or exploit security vulnerabilities. However, the company explained that the new attack went beyond that, as it was initiated, executed, and concluded entirely using independent AI tools, without direct human intervention.

The company said: "The attack was driven, from beginning to end, by an independent AI proxy system."

Fortunately, the company was able to repel the attack using its own tools, employing a large linguistic model to analyze how the breach occurred and understand its path.

"Hugging Face" provides platforms that enable developers and researchers to share and work on AI models together.

It was through these same platforms that the attacker infiltrated, uploading a dataset that contained a security vulnerability that allowed him to execute code on the company's servers.

Because the attack was fully automated, it executed thousands of actions in a very short time. While the source of the attack and the type of system used remain unknown, the company confirms it is investigating whether customer data was stolen and has pledged to keep affected users informed of any developments, advising them to review their accounts for any suspicious activity.

The company first discovered the breach through its automated systems that constantly monitor security logs, and then used other artificial intelligence systems to understand the dimensions of the attack after it occurred, enabling it to reconstruct events very quickly, accomplishing in hours what would have taken days.

However, the company faced difficulties in defending itself because it could not utilize its most robust available models due to security restrictions designed to prevent their misuse. This created a disparity, as the attacker was not bound by any rules, while the company's defensive measures were constrained by these controls.


Post a Comment

Previous Post Next Post