Researchers hacked into Gemini AI and Google Cloud Console

Researchers discovered serious security flaws in Google's AI systems

Reading time icon 2 min. read


Readers help support Windows Report. We may get a commission if you buy through our links. Tooltip Icon

Read our disclosure page to find out how can you help Windows Report sustain the editorial team. Read more

Researchers hacking into Google's AI

To discover the vulnerabilities and bugs of Bard, now known as Gemini, researchers hacked the AI. However, the Gemini AI is not their only victim. After all, the team did a number on the Google Cloud Console. In addition, after presenting their discoveries, they received $50,000. Furthermore, their efforts revealed some serious security vulnerabilities that could lead to DDOS attacks, access to specific user data, and access to uploaded images.

Can an AI be hacked?

Threat actors can hack the AI by using backdoor attacks. As a result, they can steal data and harm the model. Thus, it is important to check the system thoroughly. Also, that’s why some companies like Google and Microsoft hold events and reward researchers who are willing to hack their systems to exploit their vulnerabilities.

This research team hacked the AI and discovered how to exploit one of its vulnerabilities. Afterward, they used the major flaw to access user images without authentication, verifying processes, or permission. In addition, the research team revealed that it is possible to trick Gemini into sharing user data.

According to Cyber Security News, the research team used Gemini’s optical character recognition (OCR) to disclose other information, such as emails, notes, and revenues. Furthermore, they thought of hacking into Gemini using images with embedded data. In addition, the researchers managed to find a way to bypass the Content Security Policy (CSP).

Researchers also hacked the Google Cloud Console’s AI by discovering a vulnerability in one of the API’s endpoints. Afterward, they started exploiting it with Denial-of-Service (DoS) attacks. As a result, they prevented the AI from receiving commands from other users.

In a nutshell, the team hacked into Google Cloud Console AI and Gemini AI and managed to discover some fatal vulnerabilities. Afterward, they reported the security flaws. Furthermore, Google organizes such competitions and rewards the discovery of bugs. After all, they lead to great security-related discoveries.

What are your thoughts? Should companies host hacking events frequently? Let us know in the comments.

More about the topics: Google, Google Bard, security threats

User forum

0 messages