
FBI: Artex AI Hacking Tool Breaches Networks in Under 3 Hours
By Ali Al-Montaser — Senior AI Technical Editor
The FBI issued an urgent security warning confirming that Artex, an open-source AI-powered penetration testing tool, has been used to breach financial, government, and technology networks in the US and South Korea. Leveraging LLMs from Anthropic and OpenAI, the tool moves from targeting to full compromise in under 3 hours, resulting in the theft of data from 68,000 individuals in South Korea and the breach of 2,250 ConnectWise clients. The warning highlights a new security race where automated attackers outpace human defenders, with the open-source version still widely available despite the developer's shift to a closed model.
Executive Overview
The FBI has issued an urgent security warning confirming that Artex, an open-source AI-powered penetration testing tool, is being actively used to breach networks of financial, government, and technology organizations in the United States and South Korea. According to the warning, attacks leveraging Artex can progress from initial targeting to full network compromise in under 3 hours, a speed that significantly outpaces human defenders. The tool's ability to connect to large language models (LLMs) from Anthropic and OpenAI, as well as open-source and non-US models, enables automated reasoning for network scanning, vulnerability discovery, and exploitation. Confirmed incidents include the theft of personal data from at least 68,000 individuals in South Korea and the breach of at least 2,250 ConnectWise clients, with access to their Microsoft 365 cloud environments. Despite the developer's subsequent shift to a closed-source model, the original open-source version remains widely available, posing an ongoing risk.
📊 Official Technical Specifications & Data Sheet
| Technical Aspect | Confirmed Official Data |
|---|---|
| 💰 Pricing & Usage Cost | Free open-source tool; relies on API calls to hosted LLMs from Anthropic and OpenAI, as well as open-source models. Operating costs are tied to token pricing from model providers (e.g., official Claude and GPT prices per million input/output tokens). |
| 🌐 Platforms & Availability | Available via open-source repositories; operates through APIs for Anthropic, OpenAI, open-source, and non-US models. The original open-source version remains 'widely available' despite the developer's shift to a closed model. |
| ⚡ Performance & Speed | Moves from system targeting to full compromise in under 3 hours. Attacks are 'noisy,' generating significant data traffic that may trigger defense alerts. |
| 🛡️ Security & Breach Resistance | The tool itself is not hardened: researchers discovered hundreds of Artex systems publicly exposed online without passwords, including the system used to target South Korea. Attacks exploit simple vulnerabilities: weak passwords, exposed or stolen credentials, and lack of multi-factor authentication (MFA). |
| 🧠 Context Window | Not specified as a standalone figure; the tool relies on external LLMs (Anthropic/OpenAI/open-source) and thus inherits their context limits (typically ranging from 128K to 200K tokens depending on the model invoked). |
| 🌍 Arabic Language & Regional Support | A security tool not aimed at Arabic users as a product; no official Arabic support. However, the warning covers the Middle East within the scope of global attacks and calls for reviewing defenses in Arab organizations. |
Deep-Dive Features & Architecture
According to the FBI's urgent security warning, Artex is classified as one of several open-source AI-powered penetration testing platforms originally designed to test network defenses. Its core dangerous feature is the ability to connect to large language models (LLMs) hosted by providers such as Anthropic and OpenAI, in addition to open-source and non-US models, giving attackers automated reasoning capabilities to scan networks, discover vulnerabilities, and exploit them. The tool operates by rapidly scanning the target network for weaknesses, then exploiting simple security flaws such as weak protections and exposed or stolen passwords to establish a foothold, and then propagating within the network toward databases, mailboxes, and sensitive information.
The FBI confirmed that Artex has been used in multiple attacks since its first public appearance in July 2025. In South Korea, it was used to breach several financial institutions, largely attributed to weak cyber defenses and a lack of protections such as multi-factor authentication (MFA), allowing the theft of personal data from at least 68,000 individuals. The FBI also confirmed Microsoft's earlier findings that the tool was used to breach ConnectWise, a major IT management company, and access the Microsoft 365 cloud environments of at least 2,250 clients, with data stolen from affected customers. Security firm Barracuda reported that it suffered a breach in an attack linked to Artex, where the tool found exposed credentials belonging to a third-party service that were not revoked when the contract with the provider ended, allowing access to Barracuda's credential stores; the company had to 'assume...'
Benchmark & Competitive Performance
Artex's performance benchmarks are defined by its speed and automation. The tool's ability to complete a full network compromise in under 3 hours represents a significant acceleration compared to traditional manual penetration testing, which can take days or weeks. This speed is achieved through the integration of LLMs that automate decision-making during the attack chain. However, the tool's 'noisy' nature—generating large amounts of traffic—means it may trigger security alerts, making it less stealthy than some advanced persistent threat (APT) tactics. In terms of competitive landscape, Artex is part of a growing category of AI-powered offensive tools, but its open-source availability and confirmed real-world impact distinguish it as a notable threat. The tool's reliance on external LLM APIs also introduces latency and cost dependencies, but these are outweighed by the automation benefits for attackers.
Industry Impact & Enterprise Adoption
The FBI warning underscores a shifting cybersecurity paradigm where automated, AI-driven attacks can outpace human defenders. The confirmed breaches at ConnectWise and South Korean financial institutions demonstrate that even well-resourced organizations are vulnerable when basic security measures like MFA and credential hygiene are lacking. The exposure of hundreds of Artex systems online without passwords further highlights poor operational security among attackers, but also indicates the tool's widespread adoption. For enterprises, the warning emphasizes the need to review defenses, particularly around credential management, MFA enforcement, and network monitoring for anomalous traffic. The incident also raises questions about the dual-use nature of AI penetration testing tools and the challenges of controlling open-source software once released. As the developer shifts to a closed model, the genie is out of the bottle, and organizations must adapt to a faster, more automated threat landscape.
Conclusion
The FBI's warning about Artex marks a critical moment in cybersecurity: AI-powered hacking tools are no longer theoretical but are actively causing real-world breaches at unprecedented speed. With confirmed data theft from 68,000 individuals and 2,250 ConnectWise clients, the impact is tangible. While the developer's move to closed-source may slow future enhancements, the open-source version remains a potent weapon. Enterprises must prioritize basic security hygiene and advanced threat detection to counter this evolving risk. The race between automated attackers and human defenders is accelerating, and Artex is a stark reminder that AI is reshaping the battlefield.
Media Source: TechCrunch AI | Official Company Statement: Original Source | Fact Verification & Analysis: AI Tools Oasis
Frequently Asked Questions
Artex is an open-source AI-powered penetration testing platform that first appeared publicly in July 2025. It connects to large language models from Anthropic, OpenAI, and open-source/non-US models to execute automated cyberattacks. The developer later shifted to a closed model after its use in attacks, but the open-source version remains widely available.
The FBI confirmed that Artex attacks on South Korean financial institutions resulted in the theft of personal data from at least 68,000 individuals, while at least 2,250 ConnectWise clients were breached, with access to their Microsoft 365 cloud environments.
According to the FBI warning, Artex-related attacks can move from system targeting to full compromise in under 3 hours, a speed that outpaces human defenders' response capabilities.
Targeted sectors include financial services, government, technology, defense, telecommunications and internet companies, and law firms, according to the FBI's urgent security warning.
Yes, the FBI confirmed that the previous open-source version of Artex continues to circulate and remains 'widely available' for use in cyberattacks, despite the developer's shift to a closed model to prevent further attacks.

AI Tools Oasis Team
Bringing you the latest news and analysis in the world of Artificial Intelligence with accuracy and credibility. Follow us for all updates.


