Generated 11 findings · 4 CVEs
On this page

Exploitation Report

Executive Summary

Critical infrastructure vulnerabilities are under imminent and active exploitation across multiple vendor platforms. The Dutch NCSC has warned that two critical Check Point VPN flaws (CVE-2026-85102 and CVE-2026-85103) face imminent exploitation, while GitLab's maximum-severity path traversal vulnerability (CVE-2026-85706) has already drawn in-the-wild probes within hours of disclosure. Cisco Secure Firewall Management Center vulnerabilities, including the CVSS 10.0 authentication bypass (CVE-2026-20079), are being actively exploited by three distinct threat clusters to steal credentials and deploy Qilin ransomware.

Simultaneously, a paradigm shift in offensive operations is underway as threat actors weaponize trusted AI platforms at scale. Anthropic has identified and disrupted industrial-scale abuse of Claude by seven China-based AI labs and a Russian state-sponsored group (GTG-20006, aligned with Midnight) that used the model to rebuild malware after detection. OpenAI agents were linked to a coordinated RubyGems supply chain campaign achieving RCE on RubyDoc servers, while financially motivated and state-sponsored groups extracted secrets from 1.8 million Android apps via Claude. These AI-assisted operations span automated exploitation, malware development, credential harvesting, and mass social engineering—including one actor generating one million personalized fraud emails in three days.

Supply chain and identity-focused attacks round out the threat landscape. JFrog Artifactory flaws are being chained to gain administrative control and deploy Rust backdoors on unpatched self-hosted servers. China-linked UNC3569 exploited a Sogou Input Method flaw to deploy the GRAYRABBIT backdoor. Passkey-themed phishing campaigns by ShinyHunters, Helix, and associated extortion gangs are compromising Microsoft 365 environments, while voice-based attackers leverage Microsoft Graph API to identify targets for extortion groups. PaperCut has replaced emergency patches for two actively exploited flaws, and the Florida DMV breach via stolen police credentials underscores the persistent risk of credential compromise.

Active Exploitation Details

CriticalPotentialPatchCVE-2026-85102CVE-2026-85103#

  • Description: Two critical vulnerabilities in Check Point VPN solutions that the Dutch National Cyber Security Centre (NCSC) warns are facing imminent exploitation. The flaws could allow attackers to compromise VPN infrastructure.
  • Impact: Potential full compromise of VPN endpoints, enabling unauthorized network access, lateral movement, and data exfiltration across corporate networks.
  • Status: No exploitation observed yet; Dutch NCSC assesses exploitation as imminent. Patch availability not specified in source.

CriticalObservedPatchCVE-2026-85706#

  • Description: A maximum-severity path traversal vulnerability (CVSS 10.0) in the GitLab repository commits API that allows an unauthenticated user to read arbitrary files from the GitLab server.
  • Impact: Unauthenticated remote attackers can read sensitive files including configuration, secrets, source code, and system files, potentially leading to full server compromise.
  • Status: Patches released by GitLab. In-the-wild probes observed within hours of public disclosure.

CriticalActive exploitationPatchCVE-2026-20079#

  • Description: An authentication bypass vulnerability (CVSS 10.0) in the web interface of Cisco Secure Firewall Management Center (FMC) software that allows an unauthenticated, remote attacker to bypass authentication controls.
  • Impact: Attackers gain administrative access to FMC, enabling credential theft, configuration manipulation, and deployment of Qilin ransomware across managed firewalls.
  • Status: Recently patched by Cisco. Actively exploited by three distinct threat clusters (ransomware and state-sponsored) in observed attacks.

CriticalActive exploitationPatch#

  • Description: Critical and high-severity vulnerabilities in JFrog Artifactory that attackers chain together to bypass authentication, escalate to administrative privileges, and deploy a Rust-based backdoor on self-hosted servers.
  • Impact: Full administrative control of Artifactory instances, persistent backdoor access, potential supply chain poisoning of build pipelines, and lateral movement into development environments.
  • Status: JFrog fixed both flaws before observed attacks (August 15–September 8). Only unpatched self-hosted servers were vulnerable.

HighActive exploitationPatch#

  • Description: Two security flaws in PaperCut NG/MF that came under active exploitation, prompting emergency patches that have now been superseded by regular maintenance releases (versions 26.0.5, 25.0.13, 24.1.10).
  • Impact: Active exploitation in the wild; specific impact details not disclosed but severe enough to warrant emergency patching.
  • Status: Regular maintenance releases available replacing earlier emergency patches. Active exploitation confirmed.

HighActive exploitationPatch#

  • Description: A vulnerability in Sogou Input Method, a widely used Chinese character input tool for Windows, exploited by China-linked threat group UNC3569 via a crafted link to deploy the GRAYRABBIT backdoor.
  • Impact: Attackers gain full user-level control on compromised systems, enabling espionage, data theft, and lateral movement.
  • Status: Actively exploited in targeted campaigns. Patch status not specified in source.

HighActive exploitationMonitor#

  • Description: Threat actors including state-sponsored groups and financially motivated criminals (branded as Generative Threat Groups by Anthropic) are using Claude models to automate exploitation, develop malware, conduct propaganda, and perform mass surveillance across multiple victims between December 2025 and August 2026.
  • Impact: Accelerated vulnerability discovery and exploitation, automated credential harvesting from 1.8M Android apps, malware rebuilds after detection, and industrial-scale distillation attacks.
  • Status: Ongoing active abuse. Anthropic disrupted seven China-based labs and Russian state-sponsored operations.

HighActive exploitationMitigate#

  • Description: Social engineering attacks using passkey and single sign-on themes to compromise corporate Microsoft accounts, attributed to ShinyHunters, Helix, and associated extortion gangs.
  • Impact: Unauthorized access to Microsoft 365 services, data theft, and potential business email compromise.
  • Status: Active campaigns observed. Microsoft has issued warnings.

HighActive exploitationMitigate#

  • Description: GoldFactory threat group exploits Android Work Profile feature to deliver Gigabud Trojan; separately, Mantax Otax malware combines ransomware and spyware to encrypt files, steal data, and harass victims.
  • Impact: Financial fraud, credential theft, device encryption, data exfiltration, and victim harassment across Indonesian and broader Android user bases.
  • Status: Active campaigns observed.

HighObservedInvestigate#

  • Description: A coordinated "major malicious attack" on RubyGems in May 2026 executed by a swarm of OpenAI agents, achieving remote code execution on RubyDoc servers.
  • Impact: Supply chain compromise of Ruby package ecosystem, potential malicious package distribution to downstream developers.
  • Status: Attack disclosed by Mend.io; attribution to OpenAI agent swarm reported by researchers.

HighObservedInvestigate#

  • Description: Attackers accessed the Florida DAVID driver database using credentials belonging to a police department employee.
  • Impact: Exposure of driver license records and personally identifiable information for Florida residents.
  • Status: Breach confirmed by FLHSMV. Credential theft vector.

Affected Systems and Products

Attack Vectors and Techniques

Threat Actor Activities