Google Warns AI-Powered Cyberattack Tools Are Helping Hackers Exploit Vulnerabilities at Machine Speed

Cybercriminals are increasingly using autonomous artificial intelligence tools to identify software vulnerabilities and develop attacks, reducing the need for direct human involvement, according to Google’s latest AI threat tracker report.

The report found that threat actors are deploying autonomous large language model (LLM) frameworks to monitor software updates, identify vulnerabilities that have been patched and develop attacks targeting those flaws. These activities can take place continuously, allowing attackers to move from identifying a weakness to preparing an exploit much faster.

Google Threat Intelligence Group (GTIG) said attackers are moving beyond manual prompts and using multi-agent offensive frameworks such as ARTEX, CAIRN, CyberStrikeAI and PentAGI to carry out multiple stages of cyber intrusions.

The misuse of open-source, agentic penetration-testing frameworks is also emerging as a way to automate specific tasks and stages during live attacks, the report noted.

AI Agents Are Increasing the Speed and Scale of Cyberattacks

According to Google, the growing use of autonomous AI frameworks represents a shift towards cyberattacks operating at machine speed. Threat actors can combine different attack methods and run operations at a much larger scale, even though the underlying techniques used to break into systems have not changed significantly.

The report also highlighted the use of other tools and services, including ZeroHollow, an inference service used for exploit development and LLM routing, and Agent Zero, an open-source agent framework that supports autonomous code execution and dynamic tool generation.

Google identified financially motivated threat actors and suspected espionage groups aligned with China among those using these tools.

Hackers Use LLMs to Exploit Patched Vulnerabilities

Another concern raised in the report is the use of LLM-driven workflows to automate time-consuming tasks such as vulnerability analysis and proof-of-concept (PoC) weaponisation.

This could reduce the time between a software vendor releasing a security patch and attackers developing an exploit for the underlying vulnerability.

Google said threat actors may find it more efficient to use LLMs to weaponise patched, known vulnerabilities, also called n-day vulnerabilities. Instead of examining a broad range of information manually, attackers can use AI to compare different versions of source code, including pre-release and preview versions, and analyse release notes, security research and existing PoC code related to a specific flaw.

This gives attackers a more focused way to identify weaknesses and develop exploits after a patch has been released.

LLMjacking and Underground AI Access Services

The report also pointed to the misuse of AI infrastructure for financial gain through LLMjacking attacks.

In these attacks, threat actors steal LLM credentials and access to cloud resources from compromised enterprise environments. They then resell access through underground proxy services at discounted rates.

Google identified APIMart and JetNight as examples of such services, which are also known as LLM mixers.

The growing availability of open-source agentic frameworks is making these activities easier to carry out. Such tools allow attackers to integrate new exploits, automate operations and launch more convincing phishing campaigns, reducing the time and effort required to conduct attacks.

Google Calls for Greater Use of Agentic Cyber Defence

John Hultquist, chief analyst at Google Threat Intelligence Group, said the findings do not necessarily point to hackers developing entirely new ways to break into systems. Instead, attackers are using familiar methods with greater speed and scale.

“They are still going after known, unfixed issues, just with much greater speed and scale. Instead of relying on manual effort to keep up, we will have to embrace agentic defense. Or we risk falling behind,” Hultquist said.

His comments highlight the growing need for organisations to strengthen their security operations as attackers adopt AI tools to automate parts of the attack process.

While AI is helping security teams improve threat detection and response, Google’s findings show that the same capabilities can also be used by cybercriminals to identify weaknesses, develop exploits and scale their operations. The report points to the importance of faster vulnerability management and the adoption of AI-assisted defensive systems to keep pace with increasingly automated threats.

-By Shivani Solanki

Picture of Indian Tech Times

Indian Tech Times

Leave a Reply

Your email address will not be published. Required fields are marked *