Morphisec

We empower organizations with resilient, self-healing capabilities that prevent 
and neutralize even the most advanced cyber threats.

[Morphisec @morphisec]

𝟭,𝟬𝟬𝟬+ 𝗚𝗶𝘁𝗟𝗮𝗯 𝗦𝗲𝗿𝘃𝗲𝗿𝘀 𝗦𝘁𝗶𝗹𝗹 𝗘𝘅𝗽𝗼𝘀𝗲𝗱 𝘁𝗼 𝗠𝗮𝘅-𝗦𝗲𝘃𝗲𝗿𝗶𝘁𝘆 𝗖𝗜/𝗖𝗗 𝗥𝗖𝗘 (𝗖𝗩𝗘-𝟮𝟬𝟮𝟲-𝟵𝟳𝟮𝟮𝟵) — 𝗖𝗜𝗦𝗔 𝗔𝗱𝗱𝗲𝗱 𝘁𝗼 𝗞𝗘𝗩, 𝗙𝗲𝗱𝗲𝗿𝗮𝗹 𝗗𝗲𝗮𝗱𝗹𝗶𝗻𝗲 𝗣𝗮𝘀𝘀𝗲𝗱. Over a thousand GitLab instances remain unpatched against a maximum-severity remote code execution flaw in CI/CD pipelines, even after CISA placed it on the Known Exploited Vulnerabilities catalog and the federal remediation deadline came and went.

For security teams, this is a software supply chain exposure, not just a server patching gap. An attacker who reaches a vulnerable GitLab pipeline can execute arbitrary code in the build environment — injecting malicious code into artifacts, planting backdoors in releases shipping to customers, or using pipeline credentials to pivot deeper into internal infrastructure. If your organization runs self-hosted GitLab or consumes software built on third-party GitLab instances, verify patch status now and audit recent pipeline runs for signs of unauthorized job execution or unexpected artifact modification.

[Morphisec @morphisec]

𝗗𝗢𝗝 𝗦𝗲𝗶𝘇𝗲𝘀 𝗩𝗶𝗿𝘁𝘂𝗮𝗹𝗶𝗻𝗲 — 𝗕𝘂𝗹𝗹𝗲𝘁𝗽𝗿𝗼𝗼𝗳 𝗛𝗼𝘀𝘁𝗲𝗿 𝗧𝗵𝗮𝘁 𝗦𝗲𝗿𝘃𝗲𝗱 𝗜𝗡𝗖, 𝗗𝗿𝗮𝗴𝗼𝗻𝗙𝗼𝗿𝗰𝗲, 𝗤𝗶𝗹𝗶𝗻, 𝗖𝗼𝗶𝗻𝗯𝗮𝘀𝗲 𝗖𝗮𝗿𝘁𝗲𝗹, 𝗮𝗻𝗱 𝗦𝘁𝗲𝗮𝗹𝗖. Federal authorities seized Virtualine, exposing a shared bulletproof hosting backbone that simultaneously served some of the most active ransomware and infostealer operations in circulation — INC, DragonForce, Qilin, Coinbase Cartel, and StealC all paid customers under one roof.

For enterprise defenders, this is a rare infrastructure-layer disclosure. Virtualine-hosted IPs likely appear in C2 traffic, exfiltration endpoints, and malware delivery chains tied to multiple groups at once — meaning a single blocklist update can sever overlapping threat actor infrastructure. Audit your proxy and firewall logs for Virtualine-associated ranges now; organizations that faced any of these groups in the past 18 months should treat this as a retroactive IOC enrichment opportunity.

[Morphisec @morphisec]

𝗥𝘂𝘀𝘀𝗶𝗮-𝗔𝗹𝗶𝗴𝗻𝗲𝗱 𝗚𝗼𝗿𝗯𝗮𝗴 (𝗨𝗔���-𝟬𝟮𝟮𝟲) 𝗘𝘅𝗽𝗹𝗼𝗶𝘁𝘀 𝗪𝗶𝗻𝗥𝗔𝗥 𝗙𝗹𝗮𝘄 𝘁𝗼 𝗗𝗲𝗽𝗹𝗼𝘆 𝗚𝗜𝗙𝗧𝗘𝗗𝗖𝗥𝗢𝗢𝗞 — 𝗖𝗼𝗼𝗸𝗶𝗲 & 𝗕𝗿𝗼𝘄𝘀𝗲𝗿 𝗦𝗲𝘀𝘀𝗶𝗼𝗻 𝗧𝗵𝗲𝗳𝘁 𝗮𝘁 𝗦𝗰𝗮𝗹𝗲. Russian state-aligned group UAC-0226 is exploiting an unpatched WinRAR flaw as the entry point to drop GIFTEDCROOK, an infostealer that systematically harvests session cookies and authentication tokens from Chrome, Edge, and Firefox across compromised endpoints.

Session token theft is a post-MFA problem — stolen tokens grant adversaries fully authenticated access with no credentials required, rendering MFA controls irrelevant after the fact. Security teams should treat WinRAR patch status as a priority exposure, audit browser cookie store access via EDR telemetry, and validate that endpoint controls cover in-memory credential access patterns before exfiltration occurs. This campaign is a reminder that protecting the authentication boundary increasingly means protecting the endpoint's browser runtime, not just the identity provider.

[Morphisec @morphisec]

𝟯𝗔𝗠 𝗥𝗮𝗻𝘀𝗼𝗺𝘄𝗮𝗿𝗲 𝗔𝗳𝗳𝗶𝗹𝗶𝗮𝘁𝗲 𝗗𝗲𝗽𝗹𝗼𝘆𝘀 𝗔𝗜-𝗔𝘀𝘀𝗶𝘀𝘁𝗲𝗱 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗟𝗼𝗼𝗽 — 𝟯𝟳 𝗼𝗳 𝟱𝟬 𝗠𝗮𝗰𝗵𝗶𝗻𝗲𝘀 𝗘𝗻𝗰𝗿𝘆𝗽𝘁𝗲𝗱 𝗔𝗰𝗿𝗼𝘀𝘀 𝗨𝗞 𝗦𝗰𝗵𝗼𝗼𝗹 𝗘𝘀𝘁𝗮𝘁𝗲. Bitdefender confirmed a 3AM ransomware affiliate used an AI agentic loop — issuing commands every 40–100 seconds without human input — to encrypt 37 of 50 machines across a UK school estate. This is not a proof-of-concept. AI-assisted ransomware is operational now.

This matters for every security leader: when an agentic loop runs the attack chain, the human bottleneck is gone. The 40-second command cycle isn't a faster attacker — it's a different threat model entirely. Detection playbooks built around human dwell time assume minutes or hours between decisions. That assumption is now obsolete. If your lateral movement defenses depend on catching an attacker before they iterate, you're racing a machine that never sleeps, never second-guesses, and never makes a nervous mistake. The organizations still treating AI-assisted attacks as a future-state problem are behind. This happened. It's the new baseline.

[Morphisec @morphisec]

tweet written — Ivanti 5 VPN zero-days in 3 weeks, AMTD runtime defense framing

[Morphisec @morphisec]

𝗠𝗼𝗿𝗽𝗵𝗶𝘀𝗲𝗰 𝗧𝗵𝗿𝗲𝗮𝘁 𝗕𝗿𝗶𝗲𝗳 𝗡𝗼𝘃 𝟮 — AI is now inside the infostealer: VOLTA-AUTODOXER ranks victims by value before exfiling to Telegram. Chainalysis breached by agentic techniques. C3RB3R ransomware wipes NAS backups. Ivanti hit with 5 zero-days in 3 weeks. AI provider keys stolen via Northflank.

[Morphisec @morphisec]

𝗡𝗼𝗿𝘁𝗵𝗳𝗹𝗮𝗻𝗸 𝗕𝗿𝗲𝗮𝗰𝗵: 𝗔𝘁𝘁𝗮𝗰𝗸𝗲𝗿𝘀 𝗜𝗺𝗺𝗲𝗱𝗶𝗮𝘁𝗲𝗹𝘆 𝗪𝗲𝗮𝗽𝗼𝗻𝗶𝘇𝗲 𝗦𝘁𝗼𝗹𝗲𝗻 𝗔𝗜 𝗔𝗣𝗜 𝗞𝗲𝘆𝘀 — 𝗢𝗽𝗲𝗻𝗔𝗜, 𝗔𝗻𝘁𝗵𝗿𝗼𝗽𝗶𝗰 & 𝗢𝘁𝗵𝗲𝗿 𝗣𝗿𝗼𝘃𝗶𝗱𝗲𝗿 𝗖𝗿𝗲𝗱𝗲𝗻𝘁𝗶𝗮𝗹𝘀 𝗔𝗯𝘂𝘀𝗲𝗱. A breach of cloud development platform Northflank exposed AI API keys from multiple providers including OpenAI and Anthropic. Attackers moved within minutes — either running inference at the victim's expense or leveraging stolen LLM access to accelerate their own attack tooling.

AI API keys are now first-class credential targets on par with cloud IAM keys and service account tokens. The immediate monetization pattern means dwell time between theft and impact is near zero — there is no window to detect and rotate before damage occurs. Any organization running AI agent stacks on cloud platforms carries a credential store adversaries will actively probe. Enterprises should inventory every AI API integration across their environment, enforce usage anomaly alerting and automated key rotation, and treat AI provider credentials with the same privileged-access rigor applied to cloud IAM.

[Morphisec @morphisec]

𝗜𝘃𝗮𝗻𝘁𝗶 𝗣𝗮𝘁𝗰𝗵𝗲𝘀 𝗧𝘄𝗼 𝗠𝗼𝗿𝗲 𝗖𝗼𝗻𝗻𝗲𝗰𝘁 𝗦𝗲𝗰𝘂𝗿𝗲 𝗭𝗲𝗿𝗼-𝗗𝗮𝘆𝘀 — 𝗙𝗶𝘃𝗲 𝗜𝘃𝗮𝗻𝘁𝗶 𝗩𝗣𝗡 𝗭𝗲𝗿𝗼-𝗗𝗮𝘆𝘀 𝗶𝗻 𝗧𝗵𝗿𝗲𝗲 𝗪𝗲𝗲𝗸𝘀. Ivanti has disclosed two additional actively exploited zero-days in Connect Secure, bringing the count to five VPN vulnerabilities in roughly 21 days. This is the 15th perimeter zero-day since mid-September — a pace that makes emergency patching a losing game for most enterprise security teams.

The pattern here is the real signal for defenders: attackers are cycling through perimeter appliance zero-days faster than organizations can fully remediate them. When a VPN is the entry point, breach is increasingly a question of when, not if — which means the security posture of the internal Windows endpoints those VPN sessions reach matters as much as the patch cadence on the appliance itself. Organizations still anchoring their breach-prevention strategy entirely to perimeter patching need to pressure-test what their detection and response capability looks like the moment a threat actor lands inside the network.

[Morphisec @morphisec]

𝗦𝘆𝗻𝗼𝗹𝗼𝗴𝘆 𝗪𝗮𝗿𝗻𝘀: 𝗖𝟯𝗥𝗕𝟯𝗥 𝗥𝗮𝗻𝘀𝗼𝗺𝘄𝗮𝗿𝗲 𝗦𝘂𝗿𝗴𝗲 𝗧𝗮𝗿𝗴𝗲𝘁𝗶𝗻𝗴 𝗡𝗔𝗦 𝗗𝗲𝘃𝗶𝗰𝗲𝘀 — 𝗕𝗮𝗰𝗸𝘂𝗽 𝗗𝗲𝘀𝘁𝗿𝘂𝗰𝘁𝗶𝗼𝗻 𝗕𝗲𝗳𝗼𝗿𝗲 𝗣𝗿𝗶𝗺𝗮𝗿𝘆 𝗘𝗻𝗰𝗿𝘆𝗽𝘁𝗶𝗼𝗻. C3RB3R ransomware is surging against Synology NAS devices with a deliberate kill-chain sequence: destroy backup copies first, then encrypt primary systems. The tactical ordering eliminates the fastest recovery path before victims realize the primary environment is at risk.

The key technical detail defenders need to understand: C3RB3R generates and uses its symmetric encryption keys on the Windows endpoint, not on the NAS itself. That means endpoint memory protection operating on Windows hosts can capture those keys at execution time — creating a potential recovery path for NAS-resident files even after the device is encrypted. Shadow copy deletion also originates from the Windows host, not the appliance. Organizations using network-attached storage as backup infrastructure should evaluate whether their endpoint stack has visibility into in-memory key material during ransomware execution. The honest boundary: NAS devices sit outside endpoint telemetry entirely, so protection and key capture depend entirely on the Windows host being instrumented — if ransomware runs on an unprotected endpoint, the NAS gap remains open.

[Morphisec @morphisec]

VOLTA-AUTODOXER: AI infostealer using Claude/Kimi/Gemini to build victim dossiers and rank monetization value — three-wave Telegram exfil chain.

[Morphisec @morphisec]

𝗠𝗼𝗿𝗽𝗵𝗶𝘀𝗲𝗰 𝗧𝗵𝗿𝗲𝗮𝘁 𝗜𝗻𝘁𝗲𝗹: AI-weaponized RAT feeds creds into OpenAI before deploying ransomware, a new AMSI-bypass loader targets Asian gov builds, and Volt Typhoon's self-erasing SD-WAN framework ran 4 zero-day campaigns against US critical infrastructure.

[Morphisec @morphisec]

Tweet: Volt Typhoon Silent Train / CVE-2026-20127 Cisco SD-WAN zero-day — two paragraphs, Unicode bold

[Morphisec @morphisec]

𝗚𝗼𝗼𝗴𝗹𝗲 𝗖𝗮𝘁𝗰𝗵𝗲𝘀 𝗔𝗶𝗴𝗘𝗜 𝗦𝘂𝗽𝗽𝗹𝘇 𝗖𝗵𝗮𝗶𝗻 𝗔𝘁𝘁𝗮𝗰𝗸 — 𝟭-𝟱𝟬 𝗘𝗶𝗻𝘂𝘁𝗲 𝗧𝗮𝗺𝗽𝗲𝗿𝗶𝗻𝗴 𝗪𝗶𝗻𝗱𝗼𝗿𝘀 𝗘𝗽𝗮𝗱𝗲𝗱 𝗣𝗼𝗶𝗻𝘁-𝗶𝗻-𝗧𝗶𝗺𝗲 𝗦𝗰𝗮𝗻𝗻𝗶𝗻𝗴. Mandiant's UNC7017 compromised AOMEI's update infrastructure between April and July 2026, hot-swapping legitimate packages for CROSSROAD downloader and SHINYROAD backdoor in windows as short as five to ten minutes before reverting to the authentic build — a tempo engineered specifically to fall between periodic file-integrity checks and AV scans. Confirmed victims span higher ed, IT services, and telecom across 11 countries including the US.

Trusting vendor update channels is not a security control. When an adversary controls the update server, the delivery chain looks pristine — signed, expected, routine. The detection gap this campaign exploits is the interval between point-in-time scans; organizations relying on periodic integrity checks or pre-execution AV had no meaningful coverage of what arrived in those brief windows. Enterprise security teams should assume any trusted software distribution mechanism can be transiently compromised and ask whether their endpoint controls would catch malicious execution at runtime, independent of how clean the delivery looked.

[Morphisec @morphisec]

𝗖𝗹𝗶𝗰𝗸𝗙𝗶𝘅 𝗢𝗽𝗲𝗿𝗮𝘁𝗼𝗿𝘀 𝗪𝗲𝗮𝗽𝗼𝗻𝗶𝘇𝗲 𝗮 𝗦𝗶𝗲𝗺𝗲𝗻𝘀 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗖𝗲𝗻𝘁𝗲𝗿 ��𝗟𝗟 𝗦𝗶𝗱𝗲𝗹𝗼𝗮𝗱𝗶𝗻𝗴 𝗙𝗹𝗮𝘄 𝘁𝗼 𝗗𝗿𝗼𝗽 𝗮 𝗥𝗔𝗧 — ClickFix has evolved beyond its "paste this PowerShell" social-engineering lure. A new campaign chains the fake-CAPTCHA trick into CVE-2026-110936 (CVSS v4 8.4), a DLL sideloading flaw in Siemens Software Center. The malicious DLL executes a RAT at the same privilege level as the legitimate Siemens service — and because the host process is signed, process-reputation controls pass it without question.

Enterprises running allowlisting or signed-binary trust models face a structural blind spot here: the malicious payload never touches disk as an unsigned executable — it rides inside a vendor-blessed process. This is a direct test of whether your detection stack watches what processes do at runtime rather than what they're named or who signed them. Organizations running Siemens tooling in manufacturing or OT-adjacent environments should treat any anomalous DLL loading from Software Center as an immediate investigation priority.

[Morphisec @morphisec]

𝗡𝗲𝘁𝗮𝗺𝗟𝗼𝗮𝗱𝗲𝗿: 𝗠𝘂𝗹𝘁𝗶-𝗦𝘁𝗮𝗴𝗲 .𝗡𝗘𝗧 𝗟𝗼𝗮𝗱𝗲𝗿 𝗕𝘆𝗽𝗮𝘀𝘀𝗲𝘀 𝗔𝗠𝗦𝗜 𝗼𝗻 𝗣𝘂𝗿𝗽𝗼𝘀𝗲 𝗕𝗲𝗳𝗼𝗿𝗲 𝗜𝗻𝗷𝗲𝗰𝘁𝗶𝗻𝗴 𝗜𝗻𝘁𝗼 𝗱𝗹𝗹𝗵𝗼𝘀𝘁.𝗲𝘅𝗲 — Kaspersky's GReAT found this loader hidden in a CAB archive dropped into the Windows update packages folder of a government-customized Windows 10 build in Asia. The attack chain abuses AppDomainManager injection and a BinaryFormatter deserialization bug to unpack layered, encrypted .NET and XAML payloads, deliberately neutralizes AMSI before injecting anything, then uses COM-based injection into dllhost.exe and polls a Cloudflare-fronted C2 every 22–28 hours. One victim confirmed. No attribution to any known group — this is new tradecraft, not a recycled playbook.

"We have AMSI" is not a complete answer when the loader is architected from the ground up to bypass AMSI before the injection stage ever runs. Beyond the bypass: AppDomainManager injection and COM-based injection into dllhost.exe are both in-memory runtime operations that depend on a predictable process memory layout. Defenders relying on signature-based scanning should recognize that this chain's evasion is intentional design, not an afterthought. Government and public-sector environments face compounded risk here — the Windows update packages folder is exactly the kind of trusted, lightly monitored path that staging malware is built to exploit.

[Morphisec @morphisec]

𝗡𝗜𝗞 𝗖𝟮: 𝗔𝗜-𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗲𝗱 𝗪𝗶𝗻𝗱𝗼𝘄𝘀 𝗜𝗺𝗽𝗹𝗮𝗻𝘁 𝗗𝘂𝗺𝗽𝘀 𝗖𝗿𝗲𝗱𝗲𝗻𝘁𝗶𝗮𝗹𝘀, 𝗦𝘁𝗿𝗲𝗮𝗺𝘀 𝗞𝗲𝘆𝘀𝘁𝗿𝗼𝗸𝗲𝘀, 𝗗𝗲𝗽𝗹𝗼𝘆𝘀 𝗦𝗰𝗿𝗲𝗲𝗻-𝗟𝗼𝗰𝗸𝗶𝗻𝗴 𝗥𝗮𝗻𝘀𝗼𝗺𝘄𝗮𝗿𝗲. Cisco Talos documented NIK C2, a Go-based 64-bit Windows RAT that routes harvested credentials and keystrokes directly into the OpenAI Chat Completions API for real-time AI analysis on the compromised host — no bulk exfiltration to a traditional C2 server. The same implant can deploy screen-locking ransomware on command, communicating over TLS WebSockets on port 443 with AES-256-GCM to blend into standard HTTPS traffic.

The detection gap here is the exfiltration path. Credential dumping and keystroke capture still rely on conventional OS primitives — process memory reads, keylogging hooks — but the harvested data never leaves in a shape your DLP or UEBA tools are built to flag. Instead it rides an authenticated HTTPS POST to https://t.co/ykf2Vsysdd, visually indistinguishable from a developer hitting the same endpoint. If you have no baseline for which hosts and processes should legitimately be calling the OpenAI API, you have no anomaly to detect. The credential theft is the same as it has always been; what's new is that the evidence dispersal looks exactly like normal enterprise AI usage. That is the operational problem NIK C2 creates for defenders today.

[Morphisec @morphisec]

𝗦𝗶��𝗲𝗻𝘁 𝗟𝘆𝗻𝘅 𝗘𝘀𝗽𝗶𝗼𝗻𝗮𝗴𝗲 𝗖𝗮𝗺𝗽𝗮𝗶𝗴𝗻: 𝗪𝗶𝗻𝗥𝗔𝗥 + 𝗣𝗼𝘄𝗲𝗿𝗦𝗵𝗲𝗹𝗹 + 𝗣𝗦𝟮𝗘𝗫𝗘 𝗖𝗵𝗮𝗶𝗻 𝗗𝗿𝗼𝗽𝘀 𝗖𝘂𝘀𝘁𝗼𝗺 𝗥𝗲𝘃𝗲𝗿𝘀𝗲 𝗦𝗵𝗲𝗹𝗹𝘀 𝗶𝗻 𝗖𝗜𝗦 𝗚𝗼𝘃𝗲𝗿𝗻𝗺𝗲𝗻𝘁 𝗡𝗲𝘁𝘄𝗼𝗿𝗸𝘀. Threat actor Silent Lynx (UAT-6275) is running a staged espionage operation against government ministries and diplomatic entities across former Soviet states. The kill chain: a weaponized WinRAR archive exploits CVE-2025-8088 to drop a PowerShell loader, which then uses the PS2EXE utility to compile the script into a native Windows executable before deploying a custom reverse shell for persistent access. CISA has flagged CVE-2025-8088 as actively exploited in the wild.

The PS2EXE pivot is the detail security teams should not skim past. Converting PowerShell to a compiled executable is a deliberate evasion move — it sidesteps PowerShell script-block logging, AMSI inspection, and detection rules tuned for .ps1 activity, making the payload look like a signed-binary execution event rather than a script. For enterprises, the exposure is broader than the government sector: WinRAR is pervasive in corporate environments, and many organizations have not patched CVE-2025-8088 despite its KEV listing. Any organization with supply-chain ties, contractors, or partners operating in the CIS region should treat this campaign as a lateral-threat indicator, audit WinRAR patch status across endpoints, and add PS2EXE-generated binary heuristics to their detection backlog.

[Morphisec @morphisec]

BAE Systems contractor-as-trusted-vector cyberattack tweet

[Morphisec @morphisec]

𝐂𝐄𝐑𝐓-𝐈𝐍: 𝐀𝐈 𝐀𝐌𝐄𝐍𝐓𝐒 𝐀𝐔𝐓𝐎𝐍𝐎𝐌𝐎𝐔𝐒𝐍𝐘 𝐀𝐂𝐂𝐄𝐒𝐒𝐄𝐃 𝐒𝐄𝐍𝐒𝐈𝐓𝐈𝐕𝐄 𝐁𝐔𝐒𝐈𝐍𝐄𝐒𝐒 𝐃𝐀𝐓𝐀 𝐀𝐓 𝐈𝐍𝐃𝐈𝐀𝐍 𝐎𝐑𝐌𝐀𝐍𝐈��𝐀𝐓𝐈𝐎𝐍𝐒 — 𝐙𝐄𝐑𝐎 𝐎𝐑𝐌𝐀𝐍𝐈𝐙𝐀𝐓𝐈𝐎𝐍𝐒 𝐃𝐄𝐓𝐄𝐂𝐓𝐄𝐃 𝐈𝐓. CERT-In has confirmed that AI agents autonomously exfiltrated sensitive business data from Indian enterprises and government organizations — and not a single affected organization detected the intrusion. A separate offensive campaign used open-source AI agent frameworks to compress time-to-compromise from days to hours.

For CISOs, this is the detection gap scenario that existing tooling was not built to catch. AI agents leave no traditional signature — no malware hash, no lateral movement pattern a SIEM was tuned to flag, no human behavioral baseline to deviate from. If adversarial AI agents can operate inside your environment across a full data-access cycle without triggering a single alert, your detection coverage has a blind spot that needs to be mapped and addressed now — before the next advisory names your sector.

[Morphisec @morphisec]

𝗖𝗵𝗶𝗻𝗮'𝘀 "𝗔𝗿𝗺𝗼𝗿𝘆 𝗖𝗮𝘀𝘁𝗹𝗲" 𝗘𝘅𝗽𝗹𝗼𝗶𝘁𝗲𝗱 𝗕𝗲𝘆𝗼𝗻𝗱𝗧𝗿𝘂𝘀𝘁 𝗥𝗲𝗺𝗼𝘁𝗲 𝗦𝘂𝗽𝗽𝗼𝗿𝘁 𝗮𝘀 𝗭𝗲𝗿𝗼-𝗗𝗮𝘆 𝗦𝗶𝗻𝗰𝗲 𝗔𝘂𝗴𝘂𝘀𝘁 — 𝗥𝗼𝗼𝘁 𝗼𝗻 𝗘𝘃𝗲𝗿𝘆 𝗠𝗮𝗻𝗮𝗴𝗲𝗱 𝗦𝘆𝘀𝘁𝗲𝗺, 𝗟𝗼𝗴𝘀 𝗪𝗶𝗽𝗲𝗱. China-nexus Armory Castle (Warp Panda) weaponized CVE-2026-102764 in BeyondTrust Remote Support starting August 2026 — silently holding root on every downstream managed system while wiping logs to frustrate forensics. With 1,400+ BeyondTrust appliances still internet-exposed, the blast radius is not theoretical.

PAM platforms are the master key to the enterprise — when one is flipped into an attack vector by a nation-state actor, every system it touches is compromised in a single operation. The structural risk here is that EDR solutions grant implicit trust to RMM and PAM agents, meaning adversary activity running through BeyondTrust's own processes bypasses signature-based detection entirely. Any organization treating PAM tools as outside the threat model needs to revisit that assumption now.