The FBI has issued an urgent advisory saying hackers are using an open-source AI tool called Artex to carry out autonomous cyberattacks that can sink into target systems in less than three hours. Industries ranging from financial services and government to law firms and tech companies are being hit—locally in the U.S. and globally.
What Is Artex and How It’s Being Used
Artex is an AI-driven penetration testing tool originally released as open-source. It was built for evaluating and probing network defenses, but its capabilities let attackers automate and accelerate traditional attack steps. Users can hook Artex up to hosted powerful large language models—both U.S. ones like those from OpenAI or Anthropic and non-U.S. models—and to open-source options. Once connected, it scans networks for vulnerabilities like weak passwords or exposed credentials, then abuses them to move laterally across networks toward sensitive targets like email servers or databases.
According to the FBI, these attacks are often noisy—generating large amounts of traffic that could trigger alerts—but the speed makes detection and response difficult. In several cases, what begins as a simple vulnerability assessment turns into a full breach within under three hours.
Real-World Break-Ins and Fallout
Artex has been implicated in multiple significant breaches. One series of attacks in South Korea’s financial sector exploited poor security practices—such as weak or missing multi-factor authentication—to steal personal data from some 68,000 individuals. In another incident, the tool was used against ConnectWise, an American IT firm. That breach impacted at least 2,250 of ConnectWise’s Microsoft 365 clients, allowing unauthorized access to those cloud environments. The FBI believes a group likely linked to China is behind the ConnectWise incidents.
Another major target was cybersecurity provider Barracuda. Attackers gained access through exposed credentials tied to a vendor, a permission that should have been revoked but wasn’t. Once inside, they accessed credential stores used for gathering customer data. Because the logs didn’t fully capture the extent of access, Barracuda is assuming worst-case scenarios and has reset relevant credentials across affected systems.
Although its creator later made Artex closed-source to try and limit misuse, the tool’s earlier versions remain publicly accessible and are still in circulation among attackers. Researchers also recently discovered hundreds of Artex installations exposed online—meaning they were reachable without even a password. Among them was the very Artex system used in the financial attacks in South Korea.
These tools are changing the pace of cyber warfare. The FBI warns that what used to depend on manual reconnaissance and exploitation can now be semi-automated, minimizing the window defenders have to react. Simple defensive gaps—password reuse, lack of MFA, exposed credentials—can be enough for a fast compromise.
The advisory urges organizations to tighten basic security hygiene: enforce multi-factor authentication, secure credentials, audit and revoke vendor access, and carefully monitor network exposures. Detecting anomalous activity is essential, but too little if it only begins after damage is done.
Organizations must assume these tools will only get more capable—and attackers more comfortable using them. What Artex shows is not just a single threat, but a turning point: AI-augmented hacking is ready for the mainstream. Defenders need to treat today’s “minor” gaps as tomorrow’s full compromise.