AI agent frameworks allow hackers to breach networks in hours
New open-source tools like Artex and Cairn enable automated, around the clock cyberattacks that can move from initial targeting to privileged access in under three hours.
The short version
- The FBI warned that the Artex framework was used to breach organizations worldwide, sometimes reaching privileged access in less than three hours.
- Chinese-speaking hackers used the Cairn framework to conduct 25 operations against at least 19 organizations between August and September.
- AI agents can develop attack plans, search for vulnerabilities, and adjust tactics automatically without constant human supervision.
- Attackers are also stealing access to AI models to use those stolen computing resources to power additional cyberattacks.
- Incident response firms are currently stretched to capacity, with some turning away cases due to a lack of resources.
Why it matters
These tools shift the advantage toward hackers by allowing them to run automated, persistent attacks that don't need to sleep. For organizations, this means security teams must find ways to defend against high-speed, autonomous threats that can compromise systems before humans notice.
The rise of automated hacking
A new generation of widely available AI tools is enabling hackers to launch cyberattacks around the clock. These tools are fundamentally changing the economics of cybercrime because they allow for self-adjusting attack plans. Hackers can now use these systems to find vulnerabilities and steal credentials before human defenders realize a probe is even happening.
The FBI recently warned that an AI framework called Artex is being used to attack organizations across the globe. In some instances, these intrusions progressed from initial targeting to gaining privileged access in less than three hours. This speed highlights how quickly automated agents can navigate complex corporate networks to reach sensitive data.
Several major companies have already reported being targeted by these specific frameworks. Amazon reported that hackers used Artex to steal AI model credentials, which they then used to power additional attacks. Other reports indicate that platforms like Railway and Chainstack were also hit by attacks involving the Artex framework.
Frameworks in the wild
Beyond Artex, another open-source framework known as Cairn is also seeing significant use by malicious actors. Palo Alto Networks reported that Chinese-speaking hackers used Cairn to conduct 25 operations against at least 19 different organizations between August and September. This demonstrates that multiple different frameworks are being adopted by various threat groups.
Google also identified the use of Cairn during an intrusion into the U.S. aerospace sector. Additionally, CrowdStrike observed a group using Artex to target more than 200 organizations worldwide over an 11-day period last month. These incidents show that the tools are being used against a wide variety of industries and targets.
While these frameworks are becoming more common, they are not the only source of cyber threats. Many major services like Microsoft, Apple, and the National Security Agency have reported outages recently that were not linked to AI. However, the specific threat of AI-enabled hacking remains a primary concern for security experts due to its speed and autonomy.
How AI agents automate attacks
Artex and Cairn work by using AI agents to automate tasks that hackers traditionally had to perform manually. A hacker provides the framework with a specific target and a clear objective, such as breaking into a company network. The AI agents then take over to develop attack plans and search for potential entry points.
These agents are designed to be persistent and adaptive. They use hacking tools to test various possibilities and share findings with one another to adjust tactics when a specific move fails. This allows the software to operate continuously without needing constant human supervision to guide every single step of the process.
The efficiency of these tools is further enhanced by hackers stealing the computing resources needed to run them. By stealing access to existing AI models, attackers can expand their operations without having to pay for the underlying computing power themselves. This creates a cycle where stolen resources are used to facilitate even more theft and infiltration.
Impact on security and defense
The rapid adoption of these tools is putting a massive strain on the cybersecurity industry. John Hultquist, a chief analyst at Google, noted that incident response firms are currently stretched to capacity. Some of these firms are even turning away new cases because they simply do not have enough resources to handle the volume of attacks.
Security experts note that while these autonomous attacks generate a lot of suspicious activity, they might not always be easy to hide. However, financially motivated hackers may only need to steal valuable information quickly before defenders can intervene. This creates a high-pressure environment where the goal is to strike before a human can react.
There is also a concern regarding state-backed actors using these capabilities for more than just financial gain. Experts warn that these tools could be used to disrupt critical systems before defenders have a chance to intervene. As open-weight models reach frontier-level capabilities, the combination with tools like Artex and Cairn is hitting organizations very hard.
The path forward for defense
Because these frameworks are open-source, they are free for anyone to download and use. This accessibility means that the barrier to entry for launching sophisticated, automated attacks has been significantly lowered. Security teams are now facing a landscape where the adversary can operate at machine speed.
To keep pace, security teams may need to fundamentally change how they defend their networks. The focus is shifting toward detecting automated patterns and securing the resources that power AI models. If attackers can use stolen AI resources to launch more attacks, protecting those specific credentials becomes a priority.
While the technology is still in its infancy, the trend toward autonomous hacking is clear. Organizations must prepare for a reality where the time between an initial probe and a full breach can be measured in hours rather than days. Staying ahead will require a shift toward more proactive and automated defense mechanisms to counter the rise of AI-enabled tools.
What to watch
- How security firms develop automated defenses to counter AI agents.
- Whether state-backed actors begin using these tools for infrastructure disruption.
- New methods for securing AI model credentials from theft.
Try it
Check your organization's current protocols for detecting 'privileged access' changes that occur in under three hours.
Sources and links
- ReportingAxios: the article we sourced
First reported by Axios. 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.