GOOGLE’S Threat Intelligence Group (GTIG) warns that AI is increasingly enabling attackers with lesser resources to operate at “nation-state-level” reach. The piece describes a shift from basic prompt injection into enterprise AI systems to the development and use of weaponised AI tooling by threat actors, including state-sponsored groups, to automate and scale operations. The result is a faster attack lifecycle and an expanded attack surface as enterprises deploy additional AI-based defences.
A concrete example cited by Google researchers shows a threat actor using an AI coding chatbot, a prompt, and agent instructions to plan, build and execute a mass credential harvesting campaign in under six hours.
The report highlights notable actors and activities, including the long-running campaign by UNC6508 targeting North American academic, medical and military research institutions, and PRC-nexus groups experimenting with AI-enabled exploitation pipelines.
Baseline examples include Basin Castle using AI tools to profile targets, translate social engineering content and troubleshoot post-exploitation commands; Calanque Ion (APT42) using gen-AI for OSINT and crafting localised pretexts; Ravine Castle (APT24) employing Gemini across the attack lifecycle and for influence operations; and Midnight Neptune (UNC1069) applying AI to cryptocurrency theft.
Since March 2026, some of these actors have compromised open-source ecosystems such as PyPI, npm and Docker Hub, and introduced multiple methods to target or exploit AI tools and development workflows, including within the Dustmaker credential stealer. Google’s defensive responses include disrupting associated projects and hardening its models against misuse, including real-time protections against model extraction. The piece concludes that AI will continue to act as a force multiplier for bad actors, even as defenders try to disrupt them.