Check Point: AI agents are already running attacks almost on their own
Criminals used AI agents to attack systems with almost no human involvement, at a cost of a few hundred dollars. What changes in your company's defense.

Criminals are using AI agents to carry out hacking attacks with almost no human involvement, according to a Check Point Research report published on Thursday (the 5th). The security company analyzed campaigns run with ARTEX, a Chinese open-source platform built so that security teams can test their own systems. In the report’s words, most of the attacks were “fully autonomous”, apart from setting the initial goal.
What the agents did
Check Point found 1,608 ARTEX dashboards active and reachable on the internet. Some were misconfigured and exposed the full history of operations, which made it possible to reconstruct the campaigns.
In one of them, between October 14 and 29, the agents attacked ten stores that sell iPhone app-signing subscriptions outside the official channels and obtained more than 300 digital certificates along with their passwords. Three passed validation and could be used to sign apps. The estimated AI cost was between US$ 700 and US$ 1,500. In another, the agents recovered more than 190,000 account credentials for a game, about 6,000 of them still valid.
Not everything worked out for the criminals: in an attempt to divert cryptocurrency, the agents reported withdrawals that never happened. GBHackers confirms the main figures.
What this changes for your company
1. No one is too small to be a target. If an attack runs on its own and costs a few hundred dollars, the criminal doesn’t need to choose a victim. They try everything they find open. The report itself warns that targeted companies should expect a high volume of attempts.
2. Know what’s exposed. Check Point’s recommendation is to test and harden the systems that are visible on the internet. That starts with an up-to-date list: website, customer systems, remote access and, above all, whatever someone put online and forgot about.
3. A defense tool is also an attack tool. ARTEX was created to protect. If your team or a vendor uses AI agents in security testing, treat those installations as sensitive systems: with restricted access and out of public reach.
In 29 years of programming, I’ve seen the cost of attacking drop several times. Never this fast. Defense needs an inventory and frequent testing, not luck.
Originally published on IWA7 Insights, IWA7 Digital’s blog, which is now part of Escovando Bytes.


