Google warns of AI-powered cyberattacks using autonomous agents to scale intrusions

0
49
Google flags growing use of autonomous AI agents in cyberattacks
Google flags growing use of autonomous AI agents in cyberattacks

Cybercriminals are increasingly using autonomous AI systems to carry out attacks, steal credentials and exploit software vulnerabilities at greater speed. Google’s Threat Intelligence Group (GTIG) says attackers are moving beyond basic chatbot prompts and deploying multi-agent frameworks to automate complex cyber operations.

AI agents accelerate cyberattacks

GTIG identified tools such as ARTEX, CAIRN, CyberStrikeAI and PentAGI, which can coordinate multiple stages of an attack. These frameworks can identify vulnerabilities, generate customised payloads and adjust their methods based on results.

GTIG observed CAIRN and PentAGI being used in real attacks, while CyberStrikeAI and HexStrike AI were still being tested. CAIRN was reportedly used against several targets, including organisations in the US aerospace sector. CyberStrikeAI connects AI models with more than 100 security tools, while HexStrike AI provides access to over 150 tools managed by more than 12 specialised agents.

Between late August and early October, an attacker reportedly used a modified PentAGI framework to scan organisations across Asia. GTIG also identified a China-linked group developing its own automated attack framework.

Automated vulnerability exploitation and credential theft

GTIG reported seeing attackers automate vulnerability discovery and exploitation without human intervention for the first time. An exposed staging server hosted a tool called exploit_agent, which compared software versions before and after security patches, developed exploits and executed them.

Between January and August 2026, vulnerability disclosures and exploits nearly doubled. Monthly zero-day exploits also rose from an average of 8 in 2025 to 11 in 2026, reaching 22 in August.

The report highlighted attacks involving PaperCut and 2 PHP vulnerabilities, CVE-2026-109618 and CVE-2026-109619. In one incident, attackers reportedly gained domain administrator privileges within minutes.

Another growing threat is LLMjacking, where criminals steal AI credentials and cloud computing resources to resell access. Some incidents involved fraudulent charges exceeding $100,000 in a day, with the largest reaching $1 million.

AI also supports influence operations

GTIG said state-linked groups from Russia, China and Iran, along with commercial influence networks, are using AI for research, reconnaissance, social engineering and automated content distribution.

Attackers have also used AI-generated recruiter profiles, phishing chatbots and automated social media accounts. However, AI tools can produce inaccurate information, create excessive noise and select the wrong targets.

GTIG noted that these systems currently prioritise speed over stealth, leaving traces in network traffic, identifiers and executed code. Google SecOps has released detection rules for ARTEX, while Mandiant has published guidance to help organisations secure their systems.

Also read: Viksit Workforce for a Viksit Bharat

Do Follow: The Mainstream LinkedIn | The Mainstream Facebook | The Mainstream Youtube | The Mainstream Twitter

About us:

The Mainstream is a premier platform delivering the latest updates and informed perspectives across the technology business and cyber landscape. Built on research-driven, thought leadership and original intellectual property, The Mainstream also curates summits & conferences that convene decision makers to explore how technology reshapes industries and leadership. With a growing presence in India and globally across the Middle East, Africa, ASEAN, the USA, the UK and Australia, The Mainstream carries a vision to bring the latest happenings and insights to 8.2 billion people and to place technology at the centre of conversation for leaders navigating the future.