Tuesday, July 21, 2026
Daily digest
Actively exploited CVEs dominated the past 24 hours — Palo Alto GlobalProtect, ServiceNow AI Platform, and SonicWall SMA1000 all have confirmed in-the-wild exploitation. If any of these are in your environment, they should jump the patch queue immediately.
Critical Palo Alto GlobalProtect VPN Bug Now Exploited by Qilin Ransomware Gang
The Qilin ransomware group is actively exploiting a critical authentication bypass vulnerability in Palo Alto Networks PAN-OS GlobalProtect to breach victim networks, according to Arctic Wolf. The flaw allows unauthenticated attackers to bypass VPN authentication and gain initial network access. Patches are available from Palo Alto.
Why it matters: If GlobalProtect is part of your perimeter or remote-access stack, this is ransomware-grade initial access being actively weaponized — patch or apply mitigations before end of day and audit your GlobalProtect logs for anomalous authentication attempts.
Critical ServiceNow AI Platform Flaw (CVE-2026-6875) Exploited for Unauthenticated Code Execution
Threat actors are actively exploiting CVE-2026-6875 (CVSS 9.5), a sandbox escape vulnerability in the ServiceNow AI Platform that allows unauthenticated remote code execution. Defused Cyber confirmed in-the-wild exploitation via posts on X. ServiceNow has released patches.
Why it matters: ServiceNow is common in defense contractor and government environments for ITSM workflows — if your organization runs ServiceNow AI Platform, apply the patch immediately; unauthenticated RCE with active exploitation means this is a critical incident response priority.
SonicWall SMA1000 Flaws Exploited as Zero-Days to Deploy Custom Malware
Two recently disclosed SonicWall SMA1000 VPN appliance vulnerabilities were exploited as zero-days for weeks before disclosure, allowing attackers to install custom malware on affected devices. The flaws were used in targeted attacks against organizations relying on SMA1000 for remote access. SonicWall has since released patches.
Why it matters: Remote access appliances are high-value targets in CUI-handling environments; if SMA1000 is in use, assume potential compromise during the zero-day window and consider a forensic review of the appliance alongside patching.
HollowGraph Malware Hides C2 and Stolen Files Inside Microsoft 365 Calendar Events Dated 2050
Group-IB has documented a new espionage implant called HollowGraph that uses hijacked Microsoft 365 mailboxes as a command-and-control channel, embedding operator instructions and exfiltrated data as attachments on calendar events dated to 2050. All traffic rides the legitimate Microsoft Graph API, making it blend with normal M365 activity. The technique is designed to evade network-based detection that filters on destination or protocol.
Why it matters: Your M365 GCC High environment is directly relevant here — perimeter controls and network DLP that rely on blocking unknown destinations will not catch this technique since it abuses the Graph API over legitimate Microsoft infrastructure. Review anomaly detection rules in Defender for Office 365 and look for calendar events with far-future dates or unusual attachments.
Windows LegacyHive Zero-Day Gets Free Unofficial Patches for Privilege Escalation Flaw
A zero-day vulnerability in Windows that allows local privilege escalation on fully patched systems has been disclosed, and third-party micropatch vendor 0patch has released free unofficial patches. Microsoft has not yet issued an official fix. The flaw involves the Windows registry LegacyHive mechanism.
Why it matters: Your Intune-managed Windows 11 fleet is exposed until Microsoft ships an official patch; this is a local privilege escalation, so it matters most where endpoint security posture isn't fully hardened — monitor for Microsoft's official advisory and evaluate whether the 0patch interim fix is acceptable under your change management policy.
Microsoft Issues Manual Fix for WSUS Sync Delays and Timeouts Affecting Windows Update Scans
Microsoft has published manual mitigation steps for a known issue causing WSUS servers to experience synchronization failures, scan timeouts, and delayed Windows Update delivery, a problem that has persisted for over a week. The issue affects organizations that rely on WSUS for centralized patch distribution. A permanent fix is still in development.
Why it matters: If you use WSUS to push patches to your Windows fleet, this bug may have left systems in an unpatched state for over a week — especially critical given the LegacyHive zero-day and other active exploits this cycle. Apply the manual mitigations now and verify patch compliance in Intune or your patch reporting dashboard.
New ENCFORGE Ransomware Targets AI Model Files via Langflow RCE, Linked to JADEPUFFER Operator
Sysdig researchers have linked a second attack on a Langflow AI server to the threat actor JADEPUFFER, which now deploys ENCFORGE — a Go-based ransomware built to encrypt AI-specific assets including model weights, vector indexes, and training datasets. The entry point is Langflow's previously disclosed RCE vulnerability. This is the same operator Sysdig documented earlier in July.
Why it matters: If your self-hosted AI stack includes Langflow or similar open orchestration frameworks, this is a direct threat to those assets — audit whether Langflow is internet-exposed, apply patches, and ensure AI model data is included in your backup and recovery scope.
FakeGit Campaign Uses 7,600 Malicious GitHub Repositories to Spread SmartLoader Malware via Fake AI Tools
Researchers discovered nearly 7,600 malicious GitHub repositories, over 800 of which impersonate AI tools or Model Context Protocol (MCP) servers, used to deliver the SmartLoader malware family in a campaign called FakeGit. The repositories use convincing READMEs, copied legitimate projects, and lookalike developer profiles to appear credible. The campaign is ongoing.
Why it matters: If your team or Ansible automation pulls dependencies or AI tools from GitHub, this supply-chain campaign could introduce malware through what looks like a legitimate MCP server or AI utility — enforce repository vetting policies and consider scanning new GitHub-sourced packages before use in your environment.
New 7-Zip Vulnerability (CVE-2026-14266) Allows Code Execution via Crafted XZ Archive
A heap-based buffer overflow in 7-Zip's XZ archive processing, tracked as CVE-2026-14266, allows an attacker to execute code in the context of the current process by supplying a crafted XZ file. Trend Micro's ZDI published details on July 15; a fix was included in 7-Zip 26.02 released June 25. The vulnerability affects all prior versions.
Why it matters: 7-Zip is widely deployed on Windows endpoints and often overlooked in patch management — verify your Intune-managed fleet is on version 26.02 or later, as opening a malicious archive (e.g., delivered via phishing or a software repository) is sufficient to trigger code execution.
Amazon GuardDuty Investigation Agent Launches in Public Preview with AI-Powered Threat Assessment
AWS has released the GuardDuty investigation agent in public preview, an AI-driven feature that automatically investigates security findings across AWS environments and condenses multi-hour triage work into minutes. The agent correlates findings across accounts and workloads to produce structured investigation summaries. It is available now in AWS regions where GuardDuty is supported.
Why it matters: For your AWS GovCloud workloads, check whether this feature is available in GovCloud regions yet — if so, it could meaningfully reduce triage time for GuardDuty findings and support your IR processes without adding third-party tooling.
WordPress WP2Shell Vulnerabilities (CVE-2026-60137 + CVE-2026-63030) Under Active Mass Exploitation
Attackers are chaining two critical WordPress vulnerabilities — CVE-2026-60137 and CVE-2026-63030 — to achieve unauthenticated remote code execution and full site compromise, a combination dubbed WP2Shell. Exploitation at scale began within days of disclosure, with mass scanning observed in early morning UTC hours on Saturday. Any unpatched WordPress installation is at risk.
AI Coding Agent Sandbox Escapes Confirmed in Cursor, Codex, Gemini CLI, and Antigravity
Security researchers demonstrated sandbox escapes in four AI coding agent platforms — Cursor, OpenAI Codex, Google Gemini CLI, and Antigravity — by having the agent write files that trusted host-side tools subsequently execute, achieving code execution on the host. Multiple CVEs were assigned; patches have been issued by the affected vendors. Google downgraded severity on two Antigravity findings.
Why it matters: If developers in your environment use any of these AI coding tools on workstations that handle CUI, a sandbox escape could result in host-level code execution — verify patched versions are deployed via Intune and consider whether AI coding agent use needs to be scoped or reviewed under your CMMC access control policies.