AI-Powered Campaign Leverages Hundreds of Agents to Exploit PaperCut Flaws, Compromising 395 Organizations Globally

A sophisticated and rapidly executed global exploitation campaign, orchestrated by a threat actor believed to be Russian-speaking, has successfully compromised at least 395 organizations by exploiting critical vulnerabilities in PaperCut NG/MF print management software. The attackers employed a novel approach, utilizing hundreds of Artificial Intelligence (AI) agents to automate the development, testing, and deployment of exploits targeting CVE-2026-81578 and CVE-2026-82078. These zero-day vulnerabilities, flagged as actively exploited earlier this month, allowed for significant breaches of sensitive organizational data.

The groundbreaking findings, detailed by cybersecurity firm GreyNoise, reveal an unprecedented level of AI integration in offensive cyber operations. The campaign, which commenced on August 31st, demonstrated an alarming speed and efficiency, with attackers leveraging a combination of OpenAI’s Codex and DeepSeek models alongside conventional offensive tools. This AI-driven methodology allowed the adversary to progress from initial vulnerability discovery to achieving Domain Administrator privileges in a matter of hours, leaving many organizations with extremely limited windows for detection and response.

The Genesis of the AI-Orchestrated Attack

The genesis of this AI-driven assault can be traced to the threat actor’s strategic use of AI agents. These agents were specifically tasked with the intricate process of building, rigorously testing, and refining exploits for the two identified PaperCut vulnerabilities. This automated development pipeline significantly accelerated the exploit creation process, bypassing the more traditional, often time-consuming, manual methods employed by human attackers.

Furthermore, the AI agents were instrumental in generating comprehensive target lists. Utilizing the Netlas internet scanning and discovery platform, the agents systematically identified vulnerable PaperCut instances across the globe. This automated reconnaissance ensured a broad and efficient scope for the attack, maximizing the potential reach of the exploitation campaign.

Scope and Impact of the Breach

The scale of the compromise is substantial, with GreyNoise data indicating that the campaign affected at least 440 PaperCut instances, directly impacting 395 distinct organizations spread across 48 countries. The nature of the intrusions varied in severity:

• Credential Harvesting: 280 victims had their credentials harvested, providing attackers with access to user accounts.
• System Secrets Acquisition: 147 organizations suffered the theft of operating system or domain secrets, potentially exposing critical infrastructure information.
• Privilege Escalation: 12 organizations experienced the attackers gaining full administrator privileges, granting them complete control over affected systems.

The educational sector emerged as the most heavily impacted, accounting for approximately half of all compromised organizations. This highlights a concerning trend of critical infrastructure and sensitive data within educational institutions being targeted. Geographically, the United States bore the brunt of the attacks, followed by the United Kingdom, France, Spain, and Canada.

Interestingly, GreyNoise observed that the threat actor had specified a list of countries to avoid, including Russia, China, Iran, Ukraine, Belarus, Moldova, Brazil, and South Africa. However, the AI agents did not consistently adhere to these exclusionary rules, suggesting potential limitations or deviations in the automated decision-making processes.

An Unprecedented Timeline of Exploitation

The most alarming aspect of this campaign is the sheer speed at which it unfolded. GreyNoise’s analysis paints a stark picture of the efficiency afforded by AI orchestration:

• Initial RCE Achievement: The adversary progressed from an empty workspace to achieving Remote Code Execution (RCE) against a live victim in just under four hours.
• Domain Admin Attainment: Full Domain Administrator privileges were secured an additional two hours later.
• Rapid Compromise: Once the full campaign was launched, the attackers managed to compromise at least 11 organizations within an astonishing 26 seconds.
• Seven-Minute Domain Admin: In one particularly egregious instance, an attack against a high school in the United States saw the adversary achieve full Domain Administrator access within a mere seven minutes from initial entry.

This compressed timeline underscores the significant challenge faced by cybersecurity defenders. Traditional incident response protocols, which often rely on manual analysis and longer detection cycles, are rendered largely ineffective against such rapid, AI-driven assaults. The window for mitigating damage and preventing data exfiltration has shrunk to an unprecedented degree.

Attack Paths and Post-Exploitation Techniques

Following the successful exploitation of the PaperCut vulnerabilities, the attackers employed a series of sophisticated post-exploitation techniques to deepen their access and exfiltrate data. GreyNoise researchers identified three distinct attack paths, all of which culminated in the use of the DCSync technique.

DCSync is a powerful post-exploitation method that allows an attacker to impersonate a domain controller and request a full dump of the NTDS.DIT file. This file contains a complete replica of the Active Directory database, including all user credentials, group memberships, and security policies. By obtaining this dump, the attackers gained access to a treasure trove of sensitive domain secrets, including password hashes, which could then be further cracked or used for lateral movement within the compromised network.

The attacker’s toolkit was extensive and comprised a potent mix of well-known and custom-developed utilities. This arsenal included:

• Ligolo-ng: A versatile and powerful remote access tool for creating VPN tunnels and managing remote shells.
• Mimikatz: A classic credential dumping tool that can extract plaintext passwords, hashes, and Kerberos tickets from memory.
• Certipy: A tool designed for abusing Active Directory Certificate Services (AD CS), often used for privilege escalation and lateral movement.
• BloodHound: A powerful tool for visualizing Active Directory attack paths and identifying misconfigurations that can be exploited.
• Rubeus: A tool for Kerberos abuse, often used to obtain service tickets and perform Golden Ticket or Silver Ticket attacks.
• Impacket: A collection of Python classes for working with network protocols, commonly used for network reconnaissance and exploitation.
• NetExec: A tool for executing commands and scripts across multiple hosts on a network, facilitating rapid lateral movement.
• Custom Rust Credential-Collection Utilities: The inclusion of custom-developed tools in Rust indicates a commitment to tailoring solutions for specific tasks and potentially evading detection.

The precise objective of the threat actor remains undetermined by GreyNoise. However, the level of access and the type of data exfiltrated strongly suggest potential motives for widespread data theft, corporate espionage, or the initiation of ransomware operations. The ability to acquire a complete NTDS.DIT dump provides attackers with the foundational elements for extensive damage or lucrative extortion.

Broader Implications and Expert Analysis

The AI-orchestrated PaperCut campaign serves as a significant inflection point in the landscape of cyber threats. It demonstrates a tangible shift from human-led attacks to AI-augmented operations, fundamentally altering the speed, scale, and sophistication of cybercrime.

"The adversary went from an empty workspace to first achieving RCE against a real victim in just under four hours, first domain admin in an additional two hours, and once the full campaign launched, compromised at least 11 organizations in 26 seconds," GreyNoise noted in their analysis. This statement succinctly captures the alarming acceleration of attack lifecycles enabled by AI.

The implications for cybersecurity defenses are profound. Traditional security measures, often designed to detect and respond to human-paced attacks, are struggling to keep pace with AI-driven operations. The ability of attackers to rapidly iterate on exploits, identify targets, and execute sophisticated post-exploitation techniques with minimal human intervention presents a formidable challenge.

This event underscores the growing importance of proactive threat hunting, continuous vulnerability management, and the adoption of AI-powered defensive solutions. Organizations must move beyond reactive security postures and embrace strategies that can anticipate and counter AI-driven threats in near real-time.

Recommendations for Organizations

In light of this widespread compromise, cybersecurity professionals and system administrators are urged to take immediate action. PaperCut Software has released emergency security updates to address CVE-2026-81578 and CVE-2026-82078. Organizations utilizing PaperCut NG/MF are strongly advised to:

• Apply Emergency Security Updates: Immediately install the patches provided by PaperCut Software to remediate the exploited vulnerabilities.
• Follow Vendor Guidance: Adhere strictly to the recommendations outlined in PaperCut’s security bulletin, which provides detailed instructions for mitigation and remediation.
• Review and Harden Configurations: Conduct a thorough review of PaperCut server configurations and associated network security settings to identify and close any potential attack vectors.
• Enhance Monitoring and Detection: Implement robust logging and monitoring solutions, particularly around PaperCut servers and Active Directory, to detect suspicious activities and unauthorized access attempts.
• Credential Hygiene: Enforce strong password policies, enable multi-factor authentication where possible, and regularly audit user and administrator accounts for any anomalies.

The report also highlights a sobering statistic from a related analysis: "Once attackers have valid credentials, only 37% of their actions are blocked." This emphasizes that even with robust initial defenses, the compromise of valid credentials can severely undermine an organization’s security posture. The Blue Report 2026, which measures defenses across numerous simulations, indicates a sharp drop in prevention effectiveness once attackers leverage legitimate credentials, making credential protection and detection of their misuse paramount.

This AI-driven campaign against PaperCut serves as a stark warning of the evolving threat landscape. The convergence of AI and cybercrime is no longer a theoretical concern but a present reality, demanding a paradigm shift in how organizations approach cybersecurity. The race to develop and deploy AI for both offensive and defensive purposes is on, and staying ahead will require continuous innovation and vigilance.