Google expects fully autonomous AI-driven cyberattacks within six months
Google Threat Intelligence Group says attackers are already using AI at scale to find and exploit vulnerabilities. Researchers expect simple attacks requiring minimal human involvement within three months and fully autonomous agent-based attacks within six months.
Image: itc.ua · Author: \u0414\u043c\u0438\u0442\u0440\u043e \u041c\u0435\u043b\u044c\u043d\u0438\u043a · source articleEditorial excerpt for news reporting
Google Threat Intelligence Group says cybercriminals are moving to agent-based workflows that automate vulnerability discovery and exploitation. The activity involves both financially motivated groups and state-backed actors.
Researchers expect simple attacks requiring minimal human involvement within three months. Fully autonomous agent-based attacks are likely to appear within six months, GTIG estimates.
Open frameworks used in attacks
Google says attackers are systematically abusing open penetration-testing packages, including ARTEX, CAIRN, CyberStrikeAI and HexStrike AI. According to the report, the tools have been used for adaptive attacks against internet-facing systems with simple vulnerabilities that large language models can readily exploit.
Examples include the use of the open-source multi-agent penetration-testing framework CAIRN against the US aerospace sector. Google also reported that suspected North Korean hackers used Claude Code to deliver malware.
UNC6836 attacks and stolen access
GTIG identified several attacks by UNC6836, which is suspected of links to cybercriminals known as Midnight Neptune. The group targeted cryptocurrency companies in North America and Europe with a modified version of Claude Code.
After initial compromise, the attackers altered the application’s settings so that it loaded the SOMBERMEME backdoor when launched. Google also said LLMjacking — stealing credentials, API keys or cloud tokens to use another party’s language models — has become a common revenue source for attackers and is causing significant losses for organisations.
Less time for defenders to patch
Google said the frameworks reduce the time defenders have to install fixes and can create false trails during investigations. The FBI warned last week that several groups have used ARTEX since August against organisations in the financial, government, technology, defence, healthcare and other sectors.
Google said AI is also now regularly used in state information operations, as well as by commercial providers and political networks.
What we know
- GTIG expects fully autonomous agent-based attacks within six months.
- Attackers are using ARTEX, CAIRN, CyberStrikeAI and HexStrike AI to automate attacks.
- UNC6836 targeted cryptocurrency companies in North America and Europe with modified Claude Code.
- The compromised application was configured to load the SOMBERMEME backdoor.
- The FBI reported ARTEX use against financial, government, technology, defence and healthcare organisations.
Trust: Single source
- The story relies on one publisher. A second independent confirmation has not been established from the cited sources.
- Sources: ITC.ua.
COMMUNITY
Discussion
Sign in to join the discussion.
Loading comments…