FBI, Google sound alarm over AI-powered cyberattacks as hackers exploit new open-source frameworks
Following the FBI's Friday warning about the use of the ARTEX, an artificial intelligence framework used in cyberattacks against organizations, Google unveiled a report in which it warns about four other frameworks that use open-source AI systems to perform cyberattacks.
The tech giant called the current situation regarding AI cybersecurity a "change in paradigm," with actors being able to perform operations at machine rhythm instead of relying on the agents for prompt-specific actions that required supervision.
According to the report, cybercriminals, state-sponsored organizations, and hacktivists are using four additional frameworks: CAIRN, CyberStrikeAI, HexStrike AI, and PentAGI, each used to perform different actions during attacks.
The FBI said that ARTEX can "accelerate intrusions from initial targeting to privileged access in under three hours," revealing that "it was used in the October ConnectWise supply-chain compromise, which exposed Microsoft 365 customer environments and involved attempted authentication against more than 2,250 downstream tenants."
"We released this advisory last week because as we head to the end of the year, it’s important for Americans to protect themselves against criminals and state actors using SI to run familiar techniques at machine speed," the FBI said after the flash warning.
What do the AI frameworks do?
According to the Google report, the main reason that attackers use these frameworks is the fact that they are open-source and easy to access.
CAIRN, the report said, was used to organize agents to perform a complete operation autonomously, from an initial parameter toward an operational objective.
The report mentioned that it was recently used in an intrusion against the aerospace sector in the United States.
CyberStrikeAI and HexStrike have a similar application, in which attackers use them as a "control panel" to control different offensive software at the same time, allowing them to perform more complex operations with fewer resources.
Finally, PentAGI is used by entry-level operators because it has a lower technical barrier, allowing them to run complex campaigns focused on reconnaissance and vulnerability scanning. One of the largest campaigns done with this software targeted several organizations across Asia, the report added.
"Given the low barrier to entry, ease of integrating new exploits, and massive scaling potential, we believe it is highly likely that threat actors will continue to use these frameworks in campaigns targeting organizations globally," Google's report warned.
"As the sophistication of these tools and underlying LLMs increases, we anticipate not only reduced human involvement in these operations, but also increased stealth and sophistication of the activity," it added.
Technology ready for cybercrime, still too early for espionage risks
Google's assessment also established that the AI tools described in the report are still too noisy and not developed enough to pursue espionage and intelligence campaigns, lowering the risk of them being used by state-sponsored organizations that aim not to be uncovered while operating.
The main deterrent is the hallucination factor, which is still present in many models focused on intelligence gathering, and which may push state actors away from using these tools.
Google does warn that financially motivated actors, who are not interested in stealth when operating, would not have a problem in using these tools, even with the noise and hallucination disadvantages.
KioskNews shows a cleaned-up reading view extracted from the publisher’s page — the original always lives on their site, not ours.