Threat actors are increasingly using artificial intelligence (AI) to automate cyberattacks, moving beyond basic prompts to autonomous systems capable of discovering and exploiting vulnerabilities with limited human involvement.
According to Google Threat Intelligence Group (GTIG), both financially motivated hackers and state-sponsored groups are adopting AI-powered offensive frameworks, including ARTEX, CAIRN, CyberStrikeAI and HexStrike AI, to speed up attacks. Such tools can coordinate multiple AI agents to scan targets, identify security weaknesses and automatically attempt to exploit them.
In an advisory last week, the FBI also warned that multiple threat actors have been using the ARTEX AI pentesting framework to launch cyberattacks since August. The attacks have targeted organizations in the financial, government, technology, defense, healthcare and other sectors.
In one case, Google found a multi-agent system called exploit_agent that automated vulnerability discovery and exploitation against websites belonging to French media, entertainment and education organizations.
“Along with data exfiltrated from targeted victims, we identified multiple scripts capable of facilitating various post-compromise activities, including internal reconnaissance, brute-forcing, installing webshells, and removing other attackers’ webshells,” Google said.
Researchers warn that attacks are becoming faster and require less human involvement. Google predicts that basic attacks requiring little to no human interaction could emerge within the next three months, followed by more advanced, fully autonomous attacks within six months.
Google has also observed attackers modifying AI coding tools to deliver malware, as well as using AI in influence operations. Another growing threat is LLMjacking, where hackers compromise cloud-based AI services and use victims' computing resources to run AI models, often leaving organizations with significant bills.
“LLMJacking can generate significant usage and billing increases within hours. In some cases, single-day fraudulent charges exceeded $100,000 USD. In the most severe cases, cumulative resource abuse reached $1 million USD before remediation,” the researchers noted.
Google says that threat actors monetize LLMJacking attacks by selling access to stolen computing resources and LLM services. Researchers also noted an increase in illegal online platforms known as “LLM Mixers” that sell unauthorized access to LLM services and proxy infrastructure.
“GTIG has continued to observe threat actors leveraging LLMs to augment various phases of the attack lifecycle. However, widespread adoption of LLM tooling also introduces significant operational drawbacks. The forensically noisy nature of many LLM tool processes can degrade OPSEC and compromise stealth, creating a detrimental impact on espionage campaigns. In addition, LLMs can reduce operational reliability, as adversaries must contend with model limitations, including the hallucination of non-existent details or even targets,” the report notes.