Critical vulnerabilities, curated daily for security professionals
AI and machine-learning infrastructure dominated yesterday's disclosures, with critical flaws in IBM Langflow, vLLM, BerriAI LiteLLM, and the expr-eval JavaScript library exposing model-serving and orchestration stacks to remote compromise. The brief covers 17 critical CVEs, up 325% from the prior day's 4, and 81 high-priority CVEs, up 326% from 19. Standouts include CVE-2026-10561 (CVSS 10) in IBM Langflow OSS, CVE-2026-12866 (CVSS 9.8) in expr-eval, and CVE-2026-49468 (CVSS 9.5) in BerriAI LiteLLM, while a cluster of MISP threat-intelligence platform issues (CVE-2026-56422, CVE-2026-56423, CVE-2026-56425) adds further exposure. Remote code execution and authentication weaknesses are the recurring patterns, affecting AI pipelines, collaboration tooling, and PHP-based document processing via PhpSpreadsheet. No vendor patches were available at disclosure, so teams should prioritize compensating controls and monitor advisories for fixes.
Immediate action: Prioritize AI/ML infrastructure running IBM Langflow, vLLM, and LiteLLM, along with MISP deployments and applications bundling the expr-eval and PhpSpreadsheet libraries, and apply network restrictions or access controls where direct fixes are unavailable. With patch availability at 0%, track vendor advisories closely and immediately address the actively exploited Joomla Content Editor and Splunk issues through available workarounds or upgrades.
CVSS score (e.g. 9.1) — severity from 0–10. Red marks critical (9+), orange high (7–8.9).
Exploitability — how hard the flaw is to attack, read from the CVSS vector:
The lower the bar on all three, the easier to exploit at scale — “Network · No privileges · No interaction” is the worst case: hit from anywhere, no credentials, no victim action.
🔴 Actively exploited — confirmed under attack in the wild (CISA’s Known Exploited Vulnerabilities catalog). Prioritize these regardless of score.
EPSS · Nth percentile — FIRST.org’s estimated chance a flaw is exploited within 30 days. We flag it only in the top 10% — a statistical signal it’s unusually likely to be targeted, separate from whether attacks are confirmed.
An improper access control vulnerability in the Widget Factory Joomla Content Editor allows unauthorized users to perform restricted actions.
Splunk Enterprise and Cloud Platform contain an authentication bypass vulnerability in a PostgreSQL sidecar service, allowing unauthenticated users to create or truncate arbitrary files.
An input validation flaw in PhpSpreadsheet allows attackers to bypass wrapper protections, leading to potential Remote Code Execution (RCE) via phar deserialization.
MISP core controllers fail to properly validate ownership and primary keys, allowing authenticated users to perform unauthorized data modifications on objects they do not own.
The Azure Active Directory (AAD) authentication plugin for MISP contains multiple OAuth 2.0 implementation flaws, including session token leakage and missing HTTPS enforcement.
The expr-eval package is vulnerable to arbitrary code execution via the toJSFunction() API due to insufficient sandboxing of user-supplied expressions.
The Gaudire Assassin game contains multiple vulnerabilities in the ‘/addJugador’ endpoint, including insecure parameter handling, improper input validation, and unauthorized privilege escalation.
The Gaudire Assassin game API and local database improperly protect sensitive information, leading to the unauthorized exposure of email addresses, phone numbers, and data regarding minors.
MISP Core contains broken access-control checks in bulk deletion flows for Event Reports and Sharing Groups, allowing unauthorized deletion of data across an instance.
IBM Langflow OSS contains a critical vulnerability involving improper Python execution isolation and authentication bypass, allowing unauthenticated remote code execution.
A vulnerability in the vLLM inference engine allows unauthenticated users to bypass OpenAI API key requirements, potentially exposing LLM services to unauthorized access.
A critical vulnerability in the LiteLLM proxy server, prior to version 1.84.0, exposes the AI gateway to potential unauthorized access or exploitation.
A stored XSS vulnerability in Jupyter Server allows attackers to achieve remote code execution and gain full API authority via malicious notebook HTML payloads.
IBM Langflow OSS suffers from improper authorization enforcement in the Streamable MCP transport endpoint, enabling unauthenticated access to protected project resources and operations.
A critical flaw in the Autodesk Fusion MCP extension allows arbitrary code execution when a user visits a maliciously crafted webpage.
The Grafana Snowflake Datasource allows authenticated users to read or write arbitrary files between the local Grafana server and the Snowflake host.
An authenticated MISP administrator can trigger arbitrary code execution by providing a malicious configuration file path for the Kafka_rdkafka_config setting.
n8n contains a credential exfiltration vulnerability in the POST /rest/dynamic-node-parameters/options endpoint allowing authenticated users to bypass HTTP request domain restrictions.
Net::Statsite::Client for Perl fails to sanitize metric names and values, allowing for metric injection attacks via newline and control character manipulation.
Angular contains a security vulnerability that may allow for unauthorized operations within web applications developed on the platform.
Angular contains a security vulnerability that may affect the integrity or security of mobile and desktop web applications built using the framework.
A high-severity security vulnerability exists within the Angular development framework, potentially impacting applications built using the platform.
A high-severity security vulnerability has been identified in the Angular development platform that may affect the security posture of enterprise web applications.
A high-severity security vulnerability in the Angular framework has been disclosed, posing a risk to the security and integrity of web applications built on the platform.
A high-severity vulnerability exists in the Angular development platform that may allow for unauthorized system interaction.
A high-severity security flaw has been identified in the Angular framework that could facilitate unauthorized actions within affected web applications.
A high-severity vulnerability in the Angular development platform could potentially allow for unauthorized access or execution within the application environment.
A security vulnerability exists within the Angular development platform that may allow for unauthorized system interaction or data exposure.
A security vulnerability in the Angular development platform may expose applications to potential exploitation, requiring immediate developer attention.
A vulnerability in the Hono web application framework could potentially allow attackers to bypass security controls across various JavaScript runtimes.
A security flaw exists in the Windows Machine Config Operator (WMCO) for Red Hat OpenShift Container Platform that could lead to unauthorized system configuration or privilege escalation.
A security vulnerability in the Windows Machine Config Operator (WMCO) for Red Hat OpenShift Container Platform could allow for unauthorized configuration modifications.
A security vulnerability exists in the Vite frontend tooling framework that may allow for unauthorized system interaction.
A security vulnerability in the Filament Laravel component library may permit unauthorized access or data manipulation.
A vulnerability in the protobuf.js library may allow for malicious code execution during the compilation of protobuf definitions.
Filament, a Laravel development framework, contains a vulnerability that requires immediate attention to ensure the security of full-stack components.
MISP core contains broken access-control flaws, potentially allowing unauthorized users to perform write operations or bypass ownership checks.
The Ultimate WooCommerce Auction Pro plugin for WordPress contains an unspecified security vulnerability that could lead to unauthorized actions.
Dell Wyse Management Suite (WMS) is vulnerable to SQL injection, potentially allowing unauthenticated or authenticated attackers to execute arbitrary database commands.
The Vitepos plugin for WordPress contains a security vulnerability that may allow unauthorized access or impact the integrity of the point-of-sale system.
Dell Wyse Management Suite (WMS) contains an SQL injection vulnerability allowing unauthorized database interaction.
Apache Doris MCP Server contains a SQL injection vulnerability within its metadata query path, potentially allowing unauthorized database interactions.
The Transbank Webpay WordPress plugin contains an unspecified vulnerability in versions prior to 1 that may expose the site to unauthorized actions.
Dell Wyse Management Suite (WMS) is susceptible to an improper link resolution vulnerability, potentially leading to unauthorized file access.
The NLTK (Natural Language Toolkit) library contains a security vulnerability that requires urgent attention from developers and researchers.
A vulnerability exists in the python-multipart streaming parser that may allow for security bypasses or improper data handling.
A command injection vulnerability exists in the DHCP option processing logic of multiple TP-Link router models due to insufficient validation of externally supplied data.
SafeLine SL6 and SL6+ elevator emergency intercom systems are susceptible to an authentication bypass, potentially allowing unauthorized administrative control.
The http-proxy-middleware library for Node.js contains a vulnerability that may lead to improper proxy behavior or security bypasses.
A high-privileged remote attacker can access a hidden configuration method to modify critical program parameters within the mbCONNECT24 platform.
A critical vulnerability exists within the vLLM inference and serving engine for large language models that may lead to significant security compromise.
ArubaSign contains a vulnerability involving incorrect default permissions that may allow unauthorized access to sensitive system resources.
A high-severity vulnerability has been identified in the Nest framework, a popular tool for building scalable Node.js server-side applications.
The Angular Language Service extension for VS Code contains a high-severity vulnerability that may allow for unauthorized system interaction or code execution.
The Angular Language Service VS Code extension is susceptible to a high-severity flaw that could enable unauthorized access or execution within the development environment.
A vulnerability in MISP allows an authenticated administrator to configure an arbitrary filesystem path for the JsonLogTool error log, potentially leading to unauthorized file writes.
Crawl4AI contains an unspecified vulnerability in versions prior to 0, which may pose a significant security risk to deployments.
A Use-After-Free vulnerability exists in the xmlParseInternalSubset function of GNOME libxml2, potentially leading to memory corruption or arbitrary code execution.
A vulnerability in qSnapper before version 1 allows for the incorrect caching of authentication states between different polkit methods, potentially leading to unauthorized privilege escalation.
The qSnapper dbus service improperly caches authentication states, potentially allowing unauthorized users to access resources intended for other authenticated sessions.
An unspecified vulnerability exists in Canonical ADSys upstream versions through v0, requiring immediate review of vendor security disclosures.
A path traversal vulnerability exists in the AIL Framework, which may allow attackers to access arbitrary files on the underlying server.
A vulnerability has been identified in Yokogawa FAST/TOOLS and CI Server, which could lead to unauthorized system access or disruption of industrial control operations.
A vulnerability exists in the MessagePack-CSharp serializer that may lead to significant security impacts.
A vulnerability exists in the MessagePack-CSharp serializer that may lead to significant security impacts.
A vulnerability in Capgo versions prior to 12 poses a security risk to users of the platform.
A vulnerability exists in the protobufjs-cli command-line interface that may allow for unauthorized code execution or system compromise.
A security vulnerability exists in Picklescan, a tool used for scanning Python pickle files, which may lead to arbitrary code execution.
Picklescan contains a high-severity vulnerability that may allow for arbitrary code execution when processing malicious files.
A vulnerability exists in picklescan that may allow for malicious code execution through insecure deserialization of untrusted pickle data.
A security flaw in the piscina Node.js worker thread pool library may permit unauthorized operations or system instability.
IBM Storage Protect Client 8 contains a security vulnerability that may allow for unauthorized access or system compromise.
A time-of-check to time-of-use (TOCTOU) vulnerability exists in the polkit authentication mechanism of qSnapper, potentially allowing privilege escalation.
IBM i 7 contains a security vulnerability that may permit unauthorized system access or privilege escalation.
A stored cross-site scripting (XSS) vulnerability exists in the Runtime component of Pilz PASvisu, potentially allowing malicious script execution.
A cross-site scripting (XSS) vulnerability exists in the Builder Component of Pilz PASvisu, which could allow for unauthorized script injection.
A security vulnerability in Flowise before version 3 exists that requires immediate attention from users of the platform.
The Grafana Loki datasource plugin is vulnerable to a path traversal flaw within its callResource handler, potentially allowing unauthorized file access.
A security vulnerability exists within the vLLM inference and serving engine, potentially impacting the stability or security of LLM deployments.
A vulnerability in the MessagePack-CSharp serializer could potentially lead to security issues during the deserialization of untrusted data.
A security vulnerability exists in Capgo versions prior to 12. The specific nature of the flaw is currently under investigation by the vendor.
A high-severity vulnerability has been identified in WebP Server Go. Users are advised to verify their current version and monitor for vendor updates.
A vulnerability has been identified in the Astro web framework. Developers should review the latest security advisories to ensure their applications are protected.
An unauthenticated memory exhaustion vulnerability exists in the Grafana Enterprise public dashboard query endpoint due to missing request body size validation.
Details regarding this vulnerability in the Starlette ASGI framework remain limited, necessitating immediate review of the official vendor security advisory.
IBM i 7 contains a security vulnerability that may allow for unauthorized access or impact to system operations.
IBM WebSphere Application Server 9 contains a high-severity security vulnerability requiring immediate attention from system administrators.
Digiwin EasyFlow .NET contains an undisclosed security vulnerability that requires urgent attention and verification against vendor-provided documentation.
Chainlit versions prior to 2 contain an unspecified security vulnerability that may expose the application to unauthorized actions.
A high-severity security vulnerability has been identified in IBM WebSphere Application Server 9, necessitating prompt remediation to ensure platform security.
A security vulnerability in IBM WebSphere Application Server 9 has been disclosed, requiring urgent review and remediation to mitigate potential risks.
A security vulnerability has been identified in IBM WebSphere Application Server 8. Further details regarding the specific nature of the flaw remain under investigation.
A path traversal vulnerability in the qSnapper "configName" parameter allows attackers to potentially access unauthorized files on the underlying system.
A vulnerability in the Grafana OSS geomap panel's XYZ tile layer involves an improper ordering of sanitization and interpolation, potentially leading to security bypasses.
A malformed Bluetooth Classic SDP attribute can trigger a reachable assertion in the Zephyr RTOS SDP parser, potentially causing a denial-of-service.
A missing length validation in the Zephyr Bluetooth Host ISO receive path can be triggered by malformed HCI ISO data, leading to potential memory corruption or system crashes.
An unspecified vulnerability exists in Cap-go versions prior to 12, potentially allowing unauthorized access or impact to the application environment.
Capgo versions prior to 12 contain a vulnerability that may expose the platform to security risks. Further technical details remain limited pending vendor disclosure.
A security vulnerability has been identified within the Astro web framework. Users are advised to review the vendor's security disclosures for specific version impacts.
A security vulnerability has been identified in GitHub Copilot 1. Further details regarding the specific nature of the flaw remain under investigation.