Google exposes government-backed misuse of Gemini AI
Google's report reveals attempts by government-backed hackers to exploit Gemini AI for cyber threats, with unsuccessful jailbreak efforts reported. The company's threat intelligence department published findings indicating that attacks were aimed at using the AI chatbot for malicious purposes. Basic measures were employed in the jailbreak attempts, which were ineffective, with hackers using simple techniques like rephrasing prompts. An instance of an advanced persistent threat actor trying to use publicly available prompts to get Gemini to perform harmful coding tasks was also noted, but those attempts failed due to safety filters. Hackers from Iran, China, and North Korea were specifically reported to be engaging with Gemini for various malicious activities, including phishing campaigns, research on vulnerabilities, and exploring target network access. The report highlights a growing concern over the potential misuse of AI technologies by state-sponsored groups for cybercrime and espionage.
Source 🔗