As reported by The Hacker News, a suspected Russian-speaking threat actor has leveraged AI agents at scale to exploit a recently disclosed PaperCut NG/MF vulnerability chain, compromising over 440 instances across 395 organizations in 48 countries. This campaign — tracked independently by Blackpoint Cyber, GreyNoise, and Arctic Wolf — represents what may be the first documented large-scale operational use of AI agents to accelerate exploitation across a broad victim base.

AI Security Alert: As reported by The Hacker News, a suspected Russian-speaking threat actor has leveraged AI agents at scale to exploit a recently disclosed PaperCut NG/MF vulnerability chain, compromising over 440 instances across 395 organizations in 48 countries.

Vulnerability Details

CVETypeSeverityAffected ProductPatch AvailableActive Exploitation
CVE-2026-81578Authentication BypassCriticalPaperCut NG/MFCheck vendor advisoryYes — confirmed in the wild
CVE-2026-82078Remote Code ExecutionCriticalPaperCut NG/MFCheck vendor advisoryYes — confirmed in the wild

The chain is straightforward but devastating: an attacker bypasses authentication via CVE-2026-81578, then achieves remote code execution through CVE-2026-82078. Post-exploitation activity observed by Arctic Wolf includes Windows registry hive collection, Metasploit/Meterpreter Java payloads, and reconnaissance commands for host, user, process, and configuration enumeration.

Why This Campaign Is Different

What sets this apart from prior PaperCut exploitation waves is the adversary's operational methodology. GreyNoise reports the threat actor built a dedicated lab environment mirroring real-world PaperCut + Active Directory deployments, then orchestrated hundreds of AI agents — powered by OpenAI Codex and a DeepSeek model — to parallelize exploitation, credential harvesting, and lateral movement tooling. Public offensive security tools including Mimikatz, SharpHound, Certipy, Rubeus, and Impacket were deployed alongside these AI agents.

This is not AI writing exploit code in a vacuum. This is AI agents operating as a distributed offensive workforce — identifying targets via Netlas.io API queries, validating exploit paths, harvesting credentials, and executing post-compromise actions across hundreds of organizations simultaneously.

Who Is at Risk

  • Education sector — the primary target across the U.S., U.K., France, Spain, Canada, Belgium, Portugal, Australia, Germany, and Switzerland
  • Any organization with internet-facing PaperCut NG/MF instances, particularly those lacking MFA on admin consoles
  • Environments with Active Directory integrated into PaperCut — the lab setup and post-exploitation tooling (SharpHound, Certipy, Rubeus) indicate AD reconnaissance and potential Kerberoasting / certificate abuse paths
  • Internet-exposed print management systems in 28 countries the actor attempted to avoid — but failed to fully exclude, per GreyNoise

Broader Implications

The use of AI agents to scale exploitation changes the threat calculus. Traditionally, even capable threat actors were limited by human bandwidth — one operator can only manage so many simultaneous targets. AI agents collapse that constraint. The adversary here effectively ran a parallelized exploitation pipeline across 440+ targets with tooling that automates reconnaissance, credential harvesting, and AD mapping.

For defenders, this means the window between CVE disclosure and mass exploitation is compressing further. Organizations that treat patching as a weekly or monthly process are increasingly exposed. The PaperCut chain was recently disclosed and already weaponized at scale — defenders cannot assume any grace period.

Shield53 Recommendations

  • Patch immediately: Apply the latest PaperCut NG/MF updates addressing CVE-2026-81578 and CVE-2026-82078. If a patch is not yet available, restrict access to PaperCut admin consoles via VPN or IP allowlisting.
  • Block known IOCs: Block traffic to/from 45.142.193.132 at perimeter firewalls and review SIEM logs for any historical connections to this IP.
  • Hunt for compromise: Search for Meterpreter Java payloads, SharpHound output files, Certipy activity, and unusual PowerShell/CMD execution originating from PaperCut service accounts.
  • Enforce MFA on all PaperCut administrative interfaces and disable the built-in admin account if unused.
  • Segment print infrastructure: Ensure PaperCut servers cannot directly reach domain controllers or Tier 0 assets. AD-integrated print servers should not have privileged AD service accounts.
  • Monitor for AI-assisted recon patterns: Rapid, parallelized scanning from single IPs with varied tool signatures may indicate AI-agent-driven campaigns rather than traditional botnets.

The convergence of AI agents with publicly available offensive tooling represents an inflection point. Defenders must assume adversaries can now operate at machine speed and machine scale — and adjust detection, patching, and segmentation strategies accordingly.