LIVE FEED
FIRST LOOK Google Chrome Shifts to Two-Week Release Cycle for Faster Security Patches // FIRST LOOK Trail of Bits Ships Coop: Isolated VMs for Claude Code and Codex // CRITICAL CVE-2026-81578: PaperCut Exploited by AI Agents at Scale // FIRST LOOK arXiv Research Introduces Self-Evolving Procedural Graphs for LLM Agents // HIGH Chinese AI Firms Accused of Distilling OpenAI and Anthropic Models // HIGH Workflow Identity Hijacking Targets Enterprise AI Data Access // CRITICAL AI-Accelerated WeChat Zero-Click Worm Spreads via RCE // FIRST LOOK Microsoft Uses AI to Ship Record 974-Vulnerability Patch Batch // FIRST LOOK Meta Launches Muse Personal AI Agent with Secure VM Isolation // CRITICAL ChatGPT Cross-Account Data Leakage via Sandbox Channel //
ATLAS OWASP CRITICAL Active exploitation · Immediate action required RELEVANCE ▲ 6.2

CVE-2026-81578: PaperCut Exploited by AI Agents at Scale

TL;DR CRITICAL
  • What happened: AI agents exploiting two PaperCut flaws breached 395 orgs across 48 countries.
  • Who's at risk: PaperCut NG/MF users on unpatched versions, especially U.S. education institutions, face authentication bypass and arbitrary code execution.
  • Act now: Upgrade PaperCut NG/MF to versions 26.0.5, 25.0.13, or 24.1.10 immediately · Block or monitor traffic from IP 45.142.193[.]132 at the network perimeter · Audit PaperCut server logs for signs of unauthorised authentication or lateral movement
CVE-2026-81578: PaperCut Exploited by AI Agents at Scale

Overview

PaperCut has released full maintenance releases for its NG/MF print management software to replace a series of emergency patches addressing two actively exploited vulnerabilities: CVE-2026-81578 (authentication bypass) and CVE-2026-82078 (arbitrary code execution). Versions 26.0.5, 25.0.13, and 24.1.10 are now available and supersede all prior emergency patch releases, incorporating additional security hardening and standard QA testing.

What elevates this incident beyond a routine patch cycle is the confirmed use of AI agent infrastructure to automate exploitation at scale — a signal of a maturing AI-enabled threat landscape that warrants specific attention from the security community.

Technical Analysis

The two CVEs enable a chained attack path: CVE-2026-81578 allows an unauthenticated attacker to bypass access controls on exposed PaperCut instances, while CVE-2026-82078 enables subsequent arbitrary code execution on the compromised host.

According to GreyNoise and Blackpoint Cyber, a suspected Russian-speaking threat actor weaponised both flaws using hundreds of AI agents built on OpenAI’s Codex harness and a DeepSeek model. These agents automated target identification, exploitation sequencing, and likely evasion logic — enabling a single actor or small group to breach 395 organisations in 48 countries from a single IP address (45.142.193[.]132).

Notably, the campaign deliberately avoided targeting entities in Russia, China, Hong Kong, Thailand, Iran, and 23 other countries — a common indicator of state-adjacent threat actor tradecraft.

Framework Mapping

MITRE ATLAS:

  • AML.T0103 – Deploy AI Agent: The threat actor deployed autonomous AI agents to conduct exploitation operations at scale, replacing manual attacker workflows.
  • AML.T0047 – AI-Enabled Product or Service: OpenAI Codex and DeepSeek were used as enabling infrastructure for the attack campaign.
  • AML.T0043 – Craft Adversarial Data: Agents likely crafted and adapted exploit payloads dynamically to target diverse PaperCut configurations.

OWASP LLM Top 10:

  • LLM08 – Excessive Agency: The deployment of autonomous LLM-powered agents conducting offensive cyber operations without human-in-the-loop oversight exemplifies the systemic risk of unconstrained AI agency in adversarial hands.

Impact Assessment

The campaign’s reach — 395 organisations across 48 countries — demonstrates the force-multiplier effect of AI-assisted exploitation. The U.S. education sector bore the heaviest concentration of victims, likely due to the sector’s historically slower patch cadence and broad PaperCut adoption.

GreyNoise assesses it remains unclear whether the actor’s objective is access brokering (selling compromised footholds to ransomware affiliates) or direct follow-on action such as data theft. Either scenario carries significant downstream risk for affected institutions.

Mitigation & Recommendations

  1. Patch immediately: Upgrade to PaperCut NG/MF versions 26.0.5, 25.0.13, or 24.1.10. Emergency patch builds should be migrated to full maintenance releases.
  2. Block known malicious infrastructure: Add 45.142.193[.]132 to network-level blocklists and monitor for related traffic patterns.
  3. Review authentication logs: Inspect PaperCut server logs for anomalous login attempts, unexpected admin account creation, or unusual API activity.
  4. Restrict external exposure: Where possible, limit PaperCut admin interfaces to internal networks or VPN-only access.
  5. Monitor for follow-on activity: Given the ambiguity around actor intent, deploy endpoint detection rules for signs of lateral movement, credential dumping, or data staging post-compromise.

References

◉ AI THREAT BRIEFING

Stay ahead of the threat.

Twice-weekly digest of critical AI security developments — every story mapped to MITRE ATLAS and OWASP LLM Top 10. Free.

No spam. Unsubscribe anytime.