Google warns of the weaknesses of artificial intelligence as it targets Gemini infiltrators
Google issued a warning about the potential security risks associated with artificial intelligence (AI) after the state’s infiltrators attempted to exploit the Gemini AI model.
However, their efforts did not succeed in breaking the regime in the end.
Intelligence threat from Google Group (GTIG) has published a paper on “aggressive abuse of childbirth intelligence”, which explains how representatives of the threat spoke with Ai Chatbot, Gemini.
Blinds are trying to protect Google from Google, but they fail in simple tricks
Google stated that the actors in the threat tried to protect Gemini from artificial intelligence using claims, including the efforts made by government -backed continuous threat groups (APT) to take advantage of the artificial intelligence of malicious activities.
However, the American multinational company did not find any evidence of advanced prison attempts.
Instead, infiltrators used direct methods, such as writing different phrases or repeating the same demand several times. Google commented that these attempts did not succeed.
With the help of AI, Jailbreaks is the rapid injection attacks that try to evade restrictions and make the artificial intelligence model perform procedures, such as detecting sensitive information or creating harmful content.
According to Google, one of the appropriate actors tried to take advantage of the Georbreak claims available to the public to create a harmful symbol. However, the attempt failed, as Gemini responded to a safety candidate.
Google stated that the attackers have sought to achieve various evil goals with the help of Gemini, such as collecting information about goals, defining weaknesses from open sources, and writing code and text programs. Moreover, some attempts were made to support post -lead procedures, for example, to avoid detection.
Iran, China and North Korea exploit the Goeni Ai from Google to electronic operations
According to Google, Iran -based Apt -based APT groups are primarily raised from artificial intelligence to formulate hunting campaigns. They also used Gemini to conduct a survey of defense experts and organizations and generate content related to cybersecurity.
Meanwhile, the appropriate representatives of China have taken advantage of Gemini to explore the code, textual programming, and development tasks. They also used artificial intelligence to explore methods for deeper access to targeted networks.
The threat intelligence group from Google and North Korean groups used the appropriate Gemini in various stages of the life cycle of attack, such as research and development before the attack. The report said:
They also used Gemini to search for the strategic interest in the North Korean government, such as the South Korean army and encrypted currencies.
GTIG
Last year, North Korean infiltrators stole $ 1.3 billion of digital assets, according to the analysis.
Cryptopolitan Academy: Free Web3 cheating paper – Download now
