Threat briefs

An attacker or compromised administrator may delete organization-level IP restrictions in Anthropic to disable network-based access controls and enable the use of administrative credentials from unauthorized locations.

Administrative accounts or compromised credentials may be leveraged to perform large-scale exfiltration of organizational chat history and metadata by accessing Anthropic data export archives via signed URLs.

An automated detection framework that uses Large Language Models to correlate disparate security alerts and assess potential account compromise based on behavioral indicators.

This detection rule identifies suspicious endpoint activity, such as malicious file creation or shellcode execution, originating from or triggered by AI-assisted coding and assistant tools indicating potential supply chain or prompt injection abuse.

Detection of unauthorized GenAI workflows utilizing local compression or encoding utilities followed by outbound network communication, indicating potential staging and exfiltration of sensitive data.

Adversaries with user-level access can perform defense evasion by overwriting Claude Desktop Cowork VM boot images to execute malicious code within a virtualized guest.

ShieldCrash is a privilege escalation vulnerability in Microsoft Windows Defender that exploits a race condition during the file remediation process by leveraging symbolic link manipulation.

A command injection vulnerability in the rubygem-foreman_remote_execution component allows authenticated attackers to execute arbitrary shell commands via the Satellite API.

The Prime Mover WordPress plugin before version 2.2.1 is vulnerable to Zip Slip, allowing authenticated administrators to perform arbitrary file writes via path traversal during ZIP archive extraction.

An unauthenticated authorization bypass vulnerability in the Joomla! CMS 'com_users' component allows attackers to create accounts even when public registration is disabled.

CISA Malcolm versions prior to v26.06.0 contain multiple critical vulnerabilities, including command injection, path traversal, and SSRF, allowing attackers to achieve remote code execution, authentication bypass, and unauthorized data access.

ABB Protection and Control IED Manager PCM600 versions 2.14 and earlier contain local privilege escalation and path traversal vulnerabilities that could allow authenticated local attackers to gain unauthorized host control or overwrite files.

Multiple missing authorization vulnerabilities in the Meari IoT Cloud Platform OpenAPI Service allow authenticated users to access sensitive device data and manipulate configurations for unauthorized devices.

Multiple vulnerabilities in the Monta monta.app platform allow unauthenticated remote attackers to hijack sessions, impersonate charging stations, and disrupt services via insecure WebSocket communication.

Multiple high-severity vulnerabilities in Armatura One, including an Apache ActiveMQ deserialization flaw, expose critical physical access-control infrastructure to remote code execution and credential compromise.

Grav CMS versions 1.7.x and 2.0.0 through 2.0.24 are vulnerable to session hijacking due to an improperly restricted get_cookie() function within the Twig rendering engine.

The SVG Sanitization component in rhukster dom-sanitizer versions 1.0.15 and earlier contains an incomplete blacklist vulnerability in src/DOMSanitizer.php, allowing remote attackers to bypass security filters via malicious URL inputs.

Sharp and Toshiba multifunction printers are vulnerable to an unauthenticated path traversal attack allowing remote actors to read arbitrary sensitive system files via the installed_emanual_down.html endpoint.

InvoicePlane 1.7.1 is vulnerable to remote code execution (CVE-2026-40297) due to unsanitized input in the setup module, allowing attackers to inject arbitrary configuration directives.

Ecava IntegraXor IGX 16.0.701.10 is vulnerable to unauthenticated remote code execution via an insecure file upload endpoint that enables arbitrary command execution during service startup.

Food-Ordering 1.0 contains a Local File Inclusion (LFI) vulnerability in update_category.php allowing authenticated users to upload and execute arbitrary files via the update_image parameter.

TigerGraph Community Edition 4.2.4 contains a remote code execution chain initiated by hard-coded default credentials, enabling an arbitrary file write that allows for SSH key injection.

Teltonika RutOS 00.07.06.21 is vulnerable to post-authentication command injection via the ipsec.lua and openvpn.lua modules, allowing arbitrary root command execution.

Multiple security flaws in various OpenSSL versions allow remote attackers to perform denial of service, compromise confidentiality, and breach data integrity.

Multiple unpatched vulnerabilities in the RPM package manager allow an unauthenticated attacker to achieve arbitrary code execution on systems processing malicious packages.

A vulnerability in OpenSC, identified as CVE-2024-6559, allows a remote, authenticated attacker to execute arbitrary code through improper handling of smart card operations.

Moodle is vulnerable to several security flaws that may allow an attacker to manipulate files, perform unauthorized data disclosure, or execute cross-site scripting (XSS) attacks.

Multiple vulnerabilities in MISP allow a remote, authenticated attacker to perform privilege escalation to Site Administrator, bypass security controls, manipulate data, or execute cross-site scripting attacks.

Zammad helpdesk software is being actively exploited via two zero-day vulnerabilities, including an unauthenticated RCE (CVE-2026-102489) and an unpatched privilege escalation flaw (CVE-2026-102490).

cPanel and WHM contain multiple vulnerabilities that allow unauthenticated or authenticated attackers to perform cross-site scripting (XSS) and execute arbitrary code with administrative privileges.