Google warns that fully-automated AI attacks are coming soon
Key Points:
- Threat actors are increasingly using AI-powered agentic workflows and open-source offensive frameworks like CAIRN, ARTEX, and Claude Code to automate vulnerability discovery and exploitation, accelerating attacks to machine speed.
- Google Threat Intelligence Group (GTIG) has observed widespread adoption of AI-driven attack frameworks by both financially motivated and state-sponsored groups, with examples including attacks on the U.S. aerospace sector and cryptocurrency firms.
- GTIG warns that within three months, unsophisticated AI-driven attacks requiring minimal human involvement will become common, with fully autonomous, sophisticated AI attacks expected within six months.
- The automation of attacks reduces defenders' response time and increases false leads, as seen with the surge of AI-generated proof-of-concept exploits following recent PHP vulnerabilities, complicating patching efforts.
- Beyond automated exploitation, AI is also used in influence operations by state-backed and commercial actors, and attackers have modified AI tools like Claude Code to deploy malware such as the SOMBERMEME backdoor.