FBI warns hackers are using Artex AI to breach networks in under three hours
The FBI issued a cybersecurity advisory regarding an open-source AI tool that allows attackers to automate the discovery and exploitation of network vulnerabilities.

The short version
- Artex is an open-source AI-powered penetration testing platform used to probe network defenses.
- The tool can be connected to frontier models from Anthropic and OpenAI or various open-source models.
- Hackers used the tool to breach ConnectWise, affecting at least 2,250 Microsoft 365 cloud environments.
- The tool was also used to steal personal data from 68,000 people in South Korea's financial industry.
- The developer has closed the source code, but the FBI says the original version remains widely accessible.
Why it matters
This marks a shift toward autonomous cyberattacks where AI moves faster than human defenders can react. It highlights how easily accessible open-source tools can be weaponized to exploit simple security flaws at scale.
The rise of autonomous hacking
The FBI has issued a new cybersecurity advisory warning that hackers are increasingly relying on autonomous AI tools to launch cyberattacks. These campaigns have targeted a wide range of industries, including government, defense, tech, and financial services. The agency noted that these attacks are becoming more frequent as AI makes it easier to find and exploit common security flaws.
One specific tool identified by the FBI is Artex, an open-source AI-powered penetration testing platform. Penetration testing is a method where software is used to probe a network's defenses to find weaknesses. Because Artex is open-source, it is freely available for anyone to download and use for various purposes.
However, the FBI warns that criminals are using this accessibility to carry out widespread attacks across the United States and globally. The tool allows hackers to automate the process of identifying vulnerabilities, which significantly increases the speed of a breach. This shift toward automation means that hackers can strike multiple targets with much less manual effort than in the past.
How Artex automates breaches
Artex functions by connecting to large language models (LLMs), which are AI systems trained on vast amounts of text. Users can connect the tool to hosted frontier models from companies like Anthropic and OpenAI, or to various open-source and non-U.S. models. Once connected, the tool can be programmed to scan a victim's network rapidly to find signs of weakness.
The AI focuses on exploiting simple security bugs, such as weak protections or exposed and stolen passwords. Once it gains a foothold in a system, Artex allows the hacker to spread across the network. It can then move toward sensitive information like databases, mailboxes, and other private files belonging to the organization.
The FBI noted that while Artex intrusions can be "noisy," meaning they generate large amounts of traffic, they still move at speeds that make it difficult for human defenders to keep up. In some instances, the tool can move from the initial targeting phase to an active compromise in under three hours. This rapid progression leaves very little time for security teams to detect and stop the intrusion.
Impact on global infrastructure
Several high-profile organizations have already fallen victim to Artex-linked attacks. The FBI confirmed that the tool was used to break into ConnectWise, an IT management giant. This breach gave hackers entry into the Microsoft 365 cloud accounts of more than 2,250 of the company's clients, where they proceeded to steal data.
The tool also played a role in a major campaign against South Korea’s financial industry. These attacks targeted institutions with weak cybersecurity defenses and a lack of multi-factor authentication. As a result of these breaches, the personal data of at least 68,000 people was stolen by the attackers.
Another company, Barracuda, reported a breach where Artex found an exposed credential belonging to a third-party service. Because the credential had not been revoked after the work ended, the hackers were able to access stores of credentials used to collect data from customers. Barracuda's chief information security officer, Arve Kjoelen, noted that defenders often arrive after the attacker has finished.
The persistence of open-source risks
Following the reports of these attacks, the developer of Artex moved the software to a closed-source model. This was intended to prevent further malicious use by making the code unavailable to the public. However, the FBI warns that the previous open-source version continues to circulate and remains widely accessible for use in cyberattacks.
The risk is compounded by the fact that the tools themselves have been exposed to the public. Researchers recently discovered hundreds of publicly exposed Artex systems online that were placed on the internet without passwords. These were likely left exposed due to mistakes or misconfigurations by the people who set them up.
One of these exposed systems was the very one used to target the South Korean financial institutions. This creates a double risk: the tool can be used by bad actors to attack others, and the tool itself can be hijacked by outsiders. The FBI's warning serves as a reminder that the accessibility of AI-powered hacking tools creates a significant challenge for modern cybersecurity.
What to watch
- Whether the FBI will issue specific technical guidance on detecting Artex-specific traffic.
- How many other companies may have been affected by the ConnectWise breach beyond the 2,250 reported.
Try it
Check your organization's third-party service list to ensure all credentials for ended contracts have been revoked.
Sources and links
- Officialearlier findings from Microsoft microsoft.com
- Governmentlatest flash cybersecurity advisory ic3.gov
- ReportingTechCrunch: the article we sourced
First reported by TechCrunch. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.