AI agents exploit PaperCut: 440 servers compromised in 48 countries

The AI-driven attack that compromised 440 PaperCut servers in 48 countries

A cybercriminal suspected of speaking Russian exploited two zero-day vulnerabilities in the PaperCut NG/MF printer management software, compromising 440 instances in 48 countries. Among the victims, 204 educational institutions, including an American high school that lost complete control of the domain in just seven minutes.

Quick Response

The exploited vulnerabilities

The vulnerabilities, identified as CVE-2026-81578 and CVE-2026-82078, can be chained to allow an unauthenticated attacker to modify specific configurations and execute arbitrary Java bytecode in the security context of the PaperCut server. These vulnerabilities affect versions prior to 24.1.10, 25.0.13, and 26.0.5, depending on the installed release branch.

The technical innovation: AI agents as autonomous engineering units

The operation, partly orchestrated from the IP address 45.142.193[.]132, demonstrates how AI can accelerate already established attack techniques. The attacker employed OpenAI's Codex as an execution framework, combined with a DeepSeek model and orchestration software such as AionUI and Hindsight, a tool that provides persistent memory for AI agents.

The automation of the attack process

The AI agents did not limit themselves to generating static exploit code but functioned as an autonomous engineering unit. According to Blackpoint, the operation began in an empty workspace on August 31, 2026. The agents analyzed the patches, replicated code execution paths in a local virtual lab, built multi-thread Go-based scanning tools, and refined network probes based on real-time errors.

The impact of AI on attack speed

The strength of AI in this campaign was not an innovative exploit technique but the reduction of human effort required for research, development, debugging, classification, tracking, retries, and continuous improvement of the exploit across hundreds of real systems. Once released into public networks, the attack engine ran up to 200 concurrent threads, executing up to 100 automated retry loops.

The problem of operational instability

The incident illustrates the operational instability of agent tools. The operator attempted to impose an exclusion filter that banned intrusions in 28 nations, including Russia, China, and Iran. However, GreyNoise discovered that the exclusion controls failed in some cases, with victims identified in countries on the exclusion list, including South Africa and Brazil.

The privilege escalation techniques

The agents escalated privileges through three distinct paths: extracting LSASS process memory, abusing obsolete "noPac" defects, or exploiting Domain Controller footholds before performing complete domain credential dumps. Despite the 440 compromised instances, GreyNoise observed domain administrator access in only 12 organizations.

The challenge of post-exploit response

The campaign shows how AI-assisted workflows can drastically reduce the time and effort required to move from published vulnerability details to large-scale exploitation. It also suggests that automated attacks can generate more initial access than a human operator can immediately follow up on.

The critical defense measures

Organizations should identify PaperCut NG/MF servers accessible via the Internet and immediately update affected installations. It is crucial to apply patches for versions prior to 24.1.10, 25.0.13, and 26.0.5, depending on the installation branch. Additionally, it is essential to limit access to management interfaces and investigate exposed servers for unexpected child processes, newly created accounts, LSASS access, DCSync activity, and unusual privilege changes.

The importance of adequate cyber insurance premium

In a context where automated attacks can quickly compromise systems, it is fundamental that organizations carefully evaluate their cyber insurance policies. Adequate coverage can mitigate the financial risks associated with security breaches and ensure an effective incident response, minimizing operational disruptions and recovery costs.

The urgency of a SOC as a Service

Given the speed with which vulnerabilities can be exploited on a large scale, organizations should consider implementing a managed Security Operations Center (SOC as a Service) to continuously monitor infrastructures and respond promptly to threats. A managed SOC can provide 24/7 surveillance, detecting and countering suspicious activities before they can cause significant damage.

The need for NIS2 compliance

Automated attacks underscore the importance of NIS2 compliance. Organizations must ensure that their infrastructures are compliant with the latest regulations regarding the security of networks and information systems. This includes implementing robust security measures and conducting regular audits to identify and correct vulnerabilities.

The importance of adequate identity management

To reduce exposure to automated attacks, organizations should implement identity access management (IAM) solutions to manage and monitor user access. This includes implementing the principle of least privilege and continuously monitoring user activities to detect suspicious behaviors.

Protection against ransomware

Automated attacks can often be the starting point for more serious attacks, such as those based on ransomware. Organizations must implement ransomware protection solutions to prevent infection and ensure rapid ransomware recovery in case of an attack.