Patch-Gap Zero-Days Fuel Multi-Actor APT Surge as Management Platforms Crumble

By the Numbers

• Total articles: 50
• By severity: Critical: 14, High: 26, Low: 2, Medium: 8
• By category: APT: 2, data breach: 1, general security news: 4, malware: 8, phishing/social engineering: 3, threat actor: 5, vulnerability: 27

Top Threats

Patch-Gap Zero-Day Chain: Chrome + Windows Exploits Rapidly Adopted by Multiple Chinese APTs

The Chrome V8 type confusion (CVE-2026-85046) and WebAssembly sandbox escape (CVE-2026-87491) were fixed in Chromium source but had not yet shipped in stable browser releases, creating a window where exploit code existed but patches were unavailable to most users. At least four China-aligned groups — TA412, UTA0560, UNK_LateNight, and UNK_DoubleCheck — independently adopted the same chain within days, combining it with the Windows kernel LPE (CVE-2026-85880) to deliver ShadowPad and the GemStone credential-stealing browser extension. The speed of multi-actor adoption suggests shared infrastructure or a commercial exploit kit, and any pre-patch Chrome installation should be treated as a likely compromise vector.

• https://www.proofpoint.com/us/blog/threat-insight/once-bluemoon-multiple-state-aligned-threat-actors-rapidly-adopt-novel-exploit
• https://www.volexity.com/blog/2026/09/09/mind-the-patch-gap-multiple-chinese-threat-actors-chain-0-day-exploits-in-chrome-windows/
• https://www.cisa.gov/news-events/alerts/2026/09/09/cisa-adds-four-known-exploited-vulnerabilities-catalog
• https://research.checkpoint.com/2026/7th-september-threat-intelligence-report/

Management Platform and Perimeter Siege: RMM, Firewall, VPN, and Router Infrastructure Under Active Attack

Attackers targeted administrative tools that manage or protect thousands of downstream endpoints, exploiting CVSS 10.0 flaws in N-able N-central (CVE-2026-86218) and Cisco FMC (CVE-2026-20079) to create rogue admin accounts and gain root on firewall management consoles. Sandworm deployed Cyclops Blink implants on Cisco devices for persistent reconnaissance, while Qilin ransomware operators used legitimate FMC tooling for living-off-the-land techniques before deploying ransomware. MikroTik RouterOS, ConnectWise ScreenConnect, JFrog Artifactory, SAP NetWeaver, and Check Point Security Gateway all joined the active exploitation landscape this week, underscoring how compromise of a single management platform cascades to every device it controls.

• https://arcticwolf.com/resources/blog/cve-2026-86218/
• https://www.huntress.com/blog/n-able-vulnerability-exploitation
• https://blog.talosintelligence.com/fmc-ongoing-exploitation/
• https://www.cisa.gov/news-events/alerts/2026/09/10/cisa-adds-two-known-exploited-vulnerabilities-catalog
• https://www.cisa.gov/news-events/alerts/2026/09/11/cisa-adds-three-known-exploited-vulnerabilities-catalog
• https://www.cisa.gov/news-events/alerts/2026/09/08/cisa-adds-four-known-exploited-vulnerabilities-catalog
• https://cyber.gc.ca/en/daily-digest/2026-09-08
• https://cert.europa.eu/publications/security-advisories/2026-012/
• https://cert.europa.eu/publications/security-advisories/2026-011/

AI Systems as Attack Surface and Weapon: Distillation, Coding Assistant Exploits, and LLM Bypasses

Chinese AI companies including DeepSeek, Moonshot AI, and MiniMax conducted industrial-scale model distillation campaigns against US AI providers, sending millions of queries through proxy networks to extract reasoning and coding capabilities from Claude, GPT, and Gemini models. Separately, researchers demonstrated GitSpawn — which tricks AI coding assistants into executing malicious code before trust prompts appear — and PuzzleMask, which hides attack instructions in plain prose to bypass LLM safety gatekeepers at a 100% bypass rate. UNC6780 backdoored MCP servers with DUSTMAKER malware to steal OIDC tokens from GitHub Actions runners, showing AI infrastructure becoming both a primary target for intellectual property theft and a vector for compromising development pipelines.

• https://www.cisa.gov/news-events/cybersecurity-advisories/aa26-251a
• https://cloud.google.com/blog/topics/threat-intelligence/from-prompting-to-autonomy-the-evolution-of-adversarial-ai/
• https://research.checkpoint.com/2026/puzzlemask-abusing-plain-prose-as-a-covert-ai-attack-vector/
• https://research.checkpoint.com/2026/the-shared-clipboard-inside-the-sandbox-cross-account-data-leakage-in-chatgpt/
• https://research.checkpoint.com/2026/7th-september-threat-intelligence-report/

Browser Extension Malware and ClickFix Campaigns Target Crypto Traders and Streamers

Malicious browser extensions disguised as cryptocurrency trading tools and Twitch enhancement plugins stole session tokens and wallet data from tens of thousands of users, with J7Tracker/VREO and JeetBot exfiltrating OAuth tokens to attacker-controlled servers through browser navigation APIs that bypass CORS and CSP controls. The ClickFix social engineering technique evolved from fake CAPTCHA prompts into in-browser code execution, with SloppyRAT using finger.exe to fetch payloads and Google Visualization API for command-and-control via published Google Sheets. These campaigns operate within legitimate browser contexts and abuse trusted cloud services for exfiltration, making them difficult to detect with traditional endpoint controls.

• https://socket.dev/blog/chrome-firefox-crypto-data-theft
• https://blog.talosintelligence.com/clickfix-moves-into-the-browser/
• https://socket.dev/blog/malicious-twitch-browser-extension
• https://blog.talosintelligence.com/clearfake-webdav-infection-chain/
• https://www.zscaler.com/blogs/security-research/sloppyrat-new-tool-ransomware-attacks

Social Engineering Escalation: Passkey Vishing, AI-Powered BEC, and Targeted Phishing Lures

Storm-3121 called employees pretending to update passkeys, redirecting them to fake authentication sites that stole session tokens for cloud account takeover, then used automated tools to slowly exfiltrate files and emails while staying under 1,000 items per hour to blend with normal usage. A separate AI-assisted BEC campaign sent over one million emails impersonating CEOs and ServiceNow to finance staff, requesting fraudulent invoice payments of approximately $50,000 each with AI-generated conversation context. Meanwhile, a targeted phishing campaign against healthcare-affiliated universities used fake sexual misconduct allegations from university leadership to deliver Zoho Assist remote access tools, demonstrating how emotionally manipulative lures continue to bypass security awareness training.

• https://www.microsoft.com/en-us/security/blog/2026/09/09/passkey-themed-social-engineering-leads-identity-cloud-compromise/
• https://www.microsoft.com/en-us/security/blog/2026/09/10/protecting-organizations-ai-assisted-executive-impersonation-invoice-fraud/
• https://cofense.com/blog/false-allegations,-real-threats-sexual-misconduct-claims-used-as-phishing-lures

Trending CVEs

• CVE-2026-85880 (5 mentions) — Windows ALPC heap-based buffer overflow zero-day actively exploited in the wild for local privilege escalation to SYSTEM, used as the kernel escape component of the BlueMoon Chrome+Windows exploit chain adopted by multiple Chinese APTs. Added to CISA KEV and patched in September Patch Tuesday. Sources: 1, 2, 3, 4, 5
• CVE-2026-86218 (4 mentions) — CVSS 10.0 pre-auth static code injection in N-able N-central enabling unauthenticated remote code execution, chained with CVE-2026-86206 and CVE-2026-86207 for authentication bypass and rogue admin account creation. Actively exploited in the wild with attackers appending .invalid to email addresses to evade casual inspection. Sources: 1, 2, 3, 4
• CVE-2026-85046 (3 mentions) — Ongoing Chrome V8 JavaScript engine type confusion in the TurboFan JIT compiler enabling remote code execution in the browser renderer, exploited as zero-day in the BlueMoon chain. Was fixed in Chromium source but not yet shipped in stable Chrome releases, creating a patch-gap window. Sources: 1, 2, 3
• CVE-2026-87491 (3 mentions) — Chrome WebAssembly V8 sandbox escape via corrupting WebAssembly module metadata to execute shellcode, used as the second stage of the BlueMoon exploit chain to break out of the browser renderer sandbox. Added to CISA KEV catalog. Sources: 1, 2, 3
• CVE-2026-67277 (3 mentions) — Missing authentication for critical function in MikroTik RouterOS allowing unauthenticated device takeover, actively exploited in the wild and added to CISA KEV catalog alongside CVE-2026-86060 command injection on the same platform. Sources: 1, 2, 3
• CVE-2026-86060 (3 mentions) — Improper neutralization of argument delimiters enabling command injection on MikroTik RouterOS, actively exploited and added to CISA KEV. Combined with CVE-2026-67277, these flaws grant full unauthenticated control of router infrastructure. Sources: 1, 2, 3
• CVE-2026-20079 (2 mentions) — CVSS 10.0 unauthenticated authentication bypass on Cisco Secure Firewall Management Center granting remote root access, actively exploited by three distinct threat groups including Sandworm and Qilin ransomware operators. Sources: 1, 2
• CVE-2026-82329 (2 mentions) — Ongoing critical authentication bypass in JFrog self-hosted Artifactory (CVSS 9.8) allowing unauthenticated attackers to obtain admin access tokens, observed exploited post-disclosure on internet-exposed systems. Sources: 1, 2

Sector Trends

• Technology / Software Development — AI coding assistants, MCP servers, and LLM infrastructure became primary attack targets this week as UNC6780 backdoored development tooling to steal GitHub Actions credentials and Chinese AI companies conducted industrial-scale model distillation against US providers. GitSpawn and PuzzleMask demonstrated new classes of attacks that exploit the trust pipeline between AI gatekeepers and powerful models, making development environments a critical new frontier for defenders. Sources: 1, 2, 3, 4, 5
• Healthcare — Healthcare-affiliated universities were targeted with sexual misconduct phishing lures delivering Zoho Assist RAT, while NextGen Healthcare Mirth Connect and Orthanc DICOM Server disclosed vulnerabilities that could enable data exfiltration and denial of service in clinical environments. The convergence of social engineering and medical software vulnerabilities puts patient data and clinical operations at simultaneous risk. Sources: 1, 2, 3
• Financial / Cryptocurrency — Browser extension malware campaigns targeting crypto traders expanded across Chrome and Firefox, with J7Tracker/VREO and JeetBot stealing OAuth tokens and wallet data from over 30,000 users. An AI-assisted BEC campaign sent over one million emails impersonating executives and ServiceNow to finance staff, requesting fraudulent ACH payments of approximately $50,000 each. Sources: 1, 2, 3, 4
• Government / Defense / NGO — Chinese APT groups leveraged the BlueMoon Chrome+Windows zero-day chain to target US NGOs, aerospace firms, and government-adjacent organizations, while satellite communication terminals from ST Engineering disclosed vulnerabilities exposing authentication credentials and enabling link disruption. Sandworm's Cyclops Blink deployment on Cisco infrastructure extended state-sponsored reconnaissance capabilities into enterprise network management planes. Sources: 1, 2, 3, 4

Notable Incidents

• BlueMoon Exploit Chain: Multiple Chinese APTs Adopt Chrome+Windows Zero-Days Within Days — At least four state-aligned groups independently weaponized the same patch-gap exploit chain within days of each other, an unusually rapid multi-actor adoption pattern suggesting shared exploit infrastructure or a commercial kit. The chain combined CVE-2026-85046, CVE-2026-87491, and CVE-2026-85880 to achieve full system compromise from a single browser click.
• N-able N-central Pre-Auth RCE Actively Exploited for Downstream Endpoint Takeover — A CVSS 10.0 pre-authentication vulnerability in a widely used RMM platform let unauthenticated attackers create admin accounts and pivot to every managed endpoint, demonstrating how compromise of management infrastructure cascades at scale. Attackers appended .invalid to email addresses to evade casual admin account audits.
• Sandworm Deploys Updated Cyclops Blink on Cisco FMC Devices — Russian GRU-linked Sandworm deployed an updated Cyclops Blink variant on Cisco firewall management devices, extending the 2022 WatchGuard botnet framework with DNS-over-HTTPS C2, 80+ port scanning, and raw packet capture with content filtering. The implant masquerades as a kworker process and persists via SysV init scripts.
• CISA and Allies Expose Chinese AI Companies' Industrial-Scale Model Distillation Campaigns — A joint advisory documented how DeepSeek, Moonshot AI, MiniMax, Alibaba, StepFun, and Z.AI systematically extracted capabilities from US AI models through millions of proxied queries, using transfer stations and automated metadata sanitization to evade detection. The campaign represents a new category of intellectual property threat targeting AI providers.
• Microsoft September 2026 Patch Tuesday Sets Record with 999 Vulnerabilities — Microsoft released fixes for 999 vulnerabilities, a 70% increase over the previous monthly record, including two actively exploited zero-days (CVE-2026-85880 and CVE-2026-81963) and over 110 critical remote code execution flaws affecting DNS, RDP, DHCP, NFS, and Windows Shell.