Google Exposes Cybercriminal Efforts to Manipulate AI Chatbot Gemini

As artificial intelligence becomes more sophisticated, cybercriminals are looking for ways to manipulate it for their own purposes.

A new report from Google reveals that state-sponsored hacking groups and other threat actors have attempted to exploit its AI chatbot, Gemini, though without success.

Google’s analysis found that hackers tried to bypass Gemini’s security features using basic prompt manipulation techniques, such as rewording commands or submitting them repeatedly. While these efforts failed, the report highlights the growing interest in AI-powered cyberattacks.

Beyond simple jailbreak attempts, government-backed hacking groups have explored using Gemini for intelligence gathering, vulnerability research, and automated scripting.

Iranian cyber actors have reportedly used AI to craft phishing campaigns and monitor defense experts, while Chinese operatives have relied on it for coding assistance and deeper system access. North Korean hackers have integrated AI into their attack planning, even researching topics like cryptocurrency and military strategy.

Despite these attempts, Google confirms that Gemini’s security filters prevented any misuse. However, with North Korean hackers alone stealing $1.3 billion in digital assets in 2024, AI’s role in cyber threats is expected to grow, raising concerns about future vulnerabilities.