Skip to content
Economy Media Stay Informed Anytime, Anywhere
Tech

FBI Warns AI Hacking Tool Artex Breached Networks in Under 3 Hours

The FBI says hackers are using Artex, an AI-driven penetration testing tool, to move from first probe to full compromise in under three hours.

Computer Code
Photo: One Idea LLC via Stocksnap, CC0

The FBI has warned that hackers are turning an artificial intelligence-powered security testing tool called Artex against companies and governments in the United States and abroad, in some cases going from first contact with a target to an active compromise in “under three hours,” TechCrunch reported on Monday. The bureau’s flash advisory links the tool to attacks on a major U.S. IT management company and on financial institutions in South Korea.

What the FBI said

According to TechCrunch, the FBI’s advisory says attackers have used autonomous AI hacking tools in recent months against financial services firms, government bodies, technology and defense companies, telecommunications and internet providers, and law firms. The bureau described Artex intrusions as “noisy,” meaning they generate large volumes of network traffic that can set off security alarms. Even so, the FBI warned that the speed of these automated attacks makes it far harder for human defenders to respond in time.

The attacks described in the advisory follow a familiar pattern, just at machine speed. The software rapidly scans a victim’s network, takes advantage of basic weaknesses such as poorly protected systems or exposed and stolen passwords, and then moves sideways through the network toward databases, email inboxes and other sensitive information.

What Artex is

Artex is a penetration testing framework, a type of software that security teams use to probe their own networks for weaknesses before criminals find them. According to the cybersecurity firm Sophos, it is a multi-agent system written in the Go programming language: “planning” AI agents map a target and decide on next steps, while “worker” agents run offensive tools and report back. It records the systems and vulnerabilities it discovers and can pursue several routes into a network at once.

The tool can plug into a range of large language models, the AI systems behind chatbots. TechCrunch said that includes hosted models from companies such as Anthropic and OpenAI as well as open-source and non-U.S. models, while Sophos said operators in reported campaigns connected it to DeepSeek, GLM and Grok.

Artex was released publicly in July, according to both Sophos and TechCrunch. After it was linked to attacks, its developer made the project closed-source and took the original version offline. The FBI said the earlier open version still circulates and remains “widely accessible.” Sophos cautioned that unofficial copies should be treated as untrusted, noting that it had detected at least one backdoored version of the tool being distributed.

The attacks linked to Artex

Several incidents have been tied to the tool since its release:

  • ConnectWise: Microsoft has attributed a mid-October compromise of ConnectWise, a company that sells IT management software, to a group it tracks as Storm-0940 and assesses as aligned with the Chinese state. SecurityWeek reported that the attackers used stolen ConnectWise application credentials to reach customers’ Microsoft 365 cloud accounts, and that ConnectWise suspended several services, including Cloud Backup. Accounts of the scale differ: TechCrunch said the hackers accessed the environments of at least 2,250 customers, while SecurityWeek, citing Microsoft, said the group attempted to sign in to more than 2,250 tenants and succeeded in some. ConnectWise did not respond to TechCrunch’s request for comment.
  • South Korean finance: Several South Korean financial institutions were breached, largely because of weak defenses and a lack of protections such as multi-factor authentication, TechCrunch reported. The intrusions exposed personal data of at least 68,000 people. Sophos, citing research from CrowdStrike, dated that activity to September.
  • Barracuda: The cybersecurity company Barracuda disclosed that financially motivated criminals broke in during an Artex-linked attack using an exposed credential belonging to a former third-party vendor that had never been revoked. Because its logs were not detailed enough to rule out wider access, Barracuda reset all customer credentials, according to TechCrunch. It did not say how many customers were affected.

Microsoft said Storm-0940 had used Artex since early September against hundreds of organizations in sectors ranging from aerospace and defense to semiconductors, government and media. Describing one phase of the campaign, Microsoft said the group “used Artex to accelerate the activity to machine speed,” SecurityWeek reported. The company also said the attackers changed application credentials to keep access even after victims fixed the original problem.

Why it matters

Security researchers have long warned that AI could lower the barrier to sophisticated hacking. The Artex cases suggest that the first wave of AI-assisted attacks is less about inventing new techniques than about running old ones faster and at greater scale. Sophos described the intrusions as broadly similar to conventional attacks but with much greater scale, noise and persistence.

That also means many of the defenses are familiar. The breaches described so far relied on weak or stolen passwords, missing multi-factor authentication and credentials that were never retired. Researchers have also found hundreds of Artex systems sitting on the open internet without passwords, likely due to misconfiguration, TechCrunch reported, including the one used in the South Korean attacks.

The case is likely to add to the debate over openly releasing powerful dual-use security tools, which can help defenders find weaknesses but can just as easily be pointed at someone else’s network.