Summary: Google reports that government agents from China, Russia, Iran, and North Korea have been using its Gemini AI for malicious activities, predominantly driven by Iranian operatives. While these agents employed the AI for tasks like phishing and reconnaissance, Google claims its guardrails effectively prevented malware creation. Overall, the efforts have not led to significant advancements for threat actors, who are mostly leveraging existing capabilities of the AI for common tasks.

Affected: Google Gemini AI

Keypoints :

• Iranian agents accounted for 75% of Gemini usage among the four nations, utilizing it for phishing, vulnerability research, and cyber operations.
• Chinese operatives have used the AI primarily for content creation and researching US government institutions.
• Google’s guardrails have successfully blocked attempts to generate malware or extract personal information from the AI.
• Russian usage was limited, primarily for content manipulation related to influence campaigns.
• Google DeepMind is developing threat models and enhanced defenses to protect AI services from potential misuse.