AI Advances and the Growing Risk of Misuse: Google Threat Intelligence Report

The Dark Side of AI: Google Threat Intelligence Report Reveals Risks of Adversarial Misuse

Advances in artificial intelligence present a worrying risk of misuse since they can open doors for a variety of sectors. For instance, they can also make advances the target of hackers. In this regard, a report titled Adversarial Misuse of Generative AI, published by Google’s threat intelligence division, details how threat actors have reacted to their AI chatbot, Gemini.

Google Reports on Failed AI Jailbreak Attempts: Simple Tactics Used by Hackers

Google claimed to have observed no sophisticated attempts to use this attack vector, even if threat actors have tried to jailbreak Gemini. The goal of AI jailbreaks, also known as quick injection assaults, is to force an AI model to carry out operations that it was previously forbidden to accomplish. This includes supplying dangerous content or disclosing private information.

Hackers only employed simple tactics, such as rewording or repeatedly emailing the same prompt, according to Google. Google claimed that the attempts had failed. According to Google, in one case, an advanced persistent threat (APT) actor tricked Gemini into doing harmful coding activities by using publicly accessible prompts. However, Google said that the attempt failed because Gemini’s answer was safety-filtered.

Google Gemini AI Under Fire: APTs Use AI for Phishing, Recon, and Cyber Espionage

Along with low-effort jailbreak efforts, Google revealed that government-sponsored APTs have been targeting Gemini. Google claimed that these attackers tried to exploit Gemini to support their malevolent actions. This involved investigating known vulnerabilities, creating and scripting activities, and obtaining information about their targets.

APT actors working in Iran reportedly concentrated on leveraging AI to create phishing attacks, according to Google. They also conducted recon on defense organizations and specialists using the AI model. AI was also utilized to create cybersecurity content by the Iranian APT actors.

China’s APT actors have been using Gemini to troubleshoot development, scripting, and programming in the meantime. They also looked into ways to gain more in-depth access to their target networks using artificial intelligence. Additionally, North Korean APT actors have utilized Gemini for many stages of their attack lifecycle, including development and research.

They also used Gemini to research topics of strategic interest to the North Korean government, such as the South Korean military and cryptocurrency.

the report

For more up-to-date crypto news, you can follow Crypto Data Space.