Google researchers say hackers are using artificial
intelligence to exploit security weaknesses and are doing so with almost no
human intervention. A warning issued by Google Monday says the hackers are
using different types of AI systems to launch attacks long before a company or
person has a chance to install updates.
The warning comes from
Google’s Threat Intelligence Group (GTIG) AI Threat Tracker. A statement
provided with the report said “the era of autonomous attacks has arrived.” The
group says that attackers are using systems that have the ability to develop
malicious software, identify security flaws and exploit them automatically.
In particular, the researchers found a server hosting an AI-powered hacking system called “exploit agent.” The system examines developer updates to identify which security weaknesses were patched, then hunts for computers that haven’t installed the update yet. The hacking system can choose when and what to attack without human direction…
Key Findings from Google's Report:
Automated Patch-Hunting (exploit agent): Google
researchers discovered an exposed staging server hosting an AI-powered system
called exploit agent. The tool analyzes software developer updates to figure
out what security flaws were just patched, and then automatically hunts for and
attacks unpatched systems at machine speed. [1,
2]
Active AI Hacking Frameworks: GTIG identified other AI
systems operating in the wild, including CAIRN (which targeted organizations in
the U.S. aerospace sector) and PentAGI (which was spotted scanning systems in
Asia). [1]
Cloud Hijacking & Huge Bills: When autonomous hacking
systems break into corporate cloud accounts, they often hijack the company's
computing resources to run large-scale Large Language Models (LLMs). Some
victims have incurred more than $100,000 in fraudulent charges in a single day,
with extreme cases topping $1 million. [1]
The Speed Mismatch: John Hultquist, chief analyst for GTIG, noted that attackers are largely going after known, unfixed issues—just with unprecedented speed and scale. He warned that trying to defend against machine-speed threats using manual human patching is no longer viable, urging organizations to embrace "agentic defense" (AI-driven security systems). [1, 2]
For further coverage and details on the report, you can
read more via Cleveland.com.

No comments:
Post a Comment
Note: Only a member of this blog may post a comment.