MISP (Malware Information Sharing Platform) contains an authorization bypass in the SharingGroup model's captureSGExisting method. The authorization check used a logical AND condition that effectively exempted users with the sync permission (perm_sync) from the membership verification. Specifically, the original condition 'if (!checkIfAuthorised(...) && !perm_sync) return false' meant that any user holding the sync role would pass the check regardless of whether they were actually a member of the target sharing group.
A sync user who knows or can discover the UUID of a non-local sharing group could exploit this flaw through the /sharing_groups/add endpoint or by embedding a SharingGroup object in pushed sync data. The attacker could rewrite the sharing group record to add their own organisation as a member or insert an all_orgs server entry for their MISP instance, thereby gaining read access to all events contained in that group. This exposes threat intelligence data that was intended to be restricted to the group's legitimate members.
The vulnerability requires an authenticated account with the sync role and knowledge of a valid sharing group UUID. No user interaction is required. The impact is unauthorized disclosure of shared threat intelligence events and limited integrity modification of the sharing group record.
GitLab has remediated an issue in GitLab EE affecting all versions from 19.1 before 19.3.9, 19.4 before 19.4.8, and 19.5 before 19.5.6 that under certain conditions could have allowed an authenticated user to view unredacted secret values from secret detection findings they had access to through GitLab Duo Agent Platform due to improper authorization checks.
Attackers exploited typographical errors in domain names related to 27 AI brands to distribute malware. They used 186 typosquatting domains to deliver malicious payloads such as PowerShell commands, Microsoft Store-signed browser hijacking applications, and search-hijacking extensions affecting nearly one million users. The infrastructure includes 211 entry domains mimicking banks, insurers, and enterprise software, employing traffic profiling and platform-specific lures. This technique, termed TypoPrompting, leverages common user mistakes to redirect legitimate AI tool access into malware infection vectors.
An incorrect authorization flaw was found in AWX, the automation controller component of Red Hat Ansible Automation Platform. The API access classes for JobTemplate and WorkflowJobTemplate do not check the caller's Credential use_role when assigning the single-valued webhook_credential foreign key: JobTemplateAccess.can_add/can_change and WorkflowJobTemplateAccess.can_change in awx/main/access.py omit the check_related('webhook_credential', Credential, ..., role_field='use_role') call that UnifiedCredentialsMixin.can_attach performs for the normal many-to-many credentials relationship, and the serializer validates only the credential's type and namespace. A user who may administer their own job template (or workflow job template) and use its project and inventory can therefore attach a webhook token credential (for example a GitLab token) that they do not have use permission on. The template administrator can retrieve the template's webhook key; a subsequent webhook event causes AWX to decrypt the attached token and send it in the request header (for GitLab, PRIVATE-TOKEN) to a status URL derived from the event payload, disclosing the victim's token to a destination the attacker can influence. This may disclose another user's personal access token and allow whatever actions that token permits.
UNIX symbolic link (symlink) following vulnerability in openEuler iSulad on Linux.
This vulnerability is associated with program file src/daemon/modules/spec/specs_mount.c.
This issue affects iSulad: through 2.1.6.
If an attacker can get the victim or the platform to run their crafted malicious image, they can cause the iSulad daemon to copy sensitive host directories into an anonymous volume when creating a container, which the container can then read.
Sysdig announced Sysdig AI Defense, a runtime security and governance platform designed to monitor and control AI agents' actions in real time. The platform provides visibility into AI agent behaviors, detects risky or unauthorized actions, and can block such actions before they impact production systems. It addresses the challenge of autonomous AI agents acting faster and more complexly than human oversight can manage, offering audit-quality records and policy enforcement aligned with frameworks like the OWASP Top 10 for LLM applications.
CVE-2026-111497 is a security vulnerability in the Google Brotli decoder used by Red Hat OpenShift Dev Spaces and other Red Hat products. It involves an out-of-bounds write triggered by a crafted Brotli-compressed stream, potentially leading to memory disclosure, denial of service, or arbitrary code execution. Exploitation requires decompressing attacker-controlled Brotli data, such as HTTP content with Content-Encoding: br. Red Hat rates this flaw as Important with high attack complexity due to required conditions and platform hardenings. No fix is currently available in the Red Hat OpenShift Dev Spaces 3.31.2 RHEL 10 container release, and users are advised to migrate from devfile v1 to v2 standards and update to supported OpenShift releases. The vulnerability affects Brotli decoder usage in various Node.js and .NET versions depending on whether they use system or bundled Brotli libraries.
ServiceNow has remediated a sandbox escape security issue that was identified in the ServiceNow AI Platform. This vulnerability could enable an authenticated user with the admin role to bypass Rhino sandbox restrictions and execute arbitrary code on the underlying host operating system, potentially enabling further unintended access.
ServiceNow deployed an update to hosted instances, and ServiceNow provided the update to our partners and self-hosted customers. We are not currently aware of malicious exploitation against ServiceNow instances. We recommend customers promptly apply appropriate updates or upgrade to a patched release if they have not already done so.
ServiceNow has addressed a domain separation bypass security issue that was identified in the ServiceNow SR – CSM – Task Plan Templates Store application. This vulnerability could enable an authenticated user with a specific role to create records in domains they do not have access to, potentially leading to unauthorized data creation across domain boundaries.
We are not currently aware of malicious exploitation against ServiceNow instances. We recommend customers promptly apply appropriate updates or upgrade to a patched release if they have not already done so.
ServiceNow has remediated an information disclosure vulnerability that was identified in the ServiceNow AI Platform. This vulnerability could cause a user's plaintext password to be persisted in a system table, potentially exposing that password to a user who otherwise would not be entitled to view it.
ServiceNow remediated the vulnerability through security updates that address the unintended persistence of plaintext credentials. We are not currently aware of malicious exploitation against ServiceNow instances. We recommend customers promptly apply appropriate updates or upgrade to a patched release if they have not already done so.
Several Russian-language traffer teams have adapted affiliate-fraud models to exploit the growing popularity of prediction markets such as PolyMarket and Kalshi. These teams provide full fraud infrastructure including fake websites, admin panels, cryptocurrency payment systems, and support. They use controlled trading outcomes, fabricated market activity, artificial withdrawal restrictions, and collect sensitive identity information. Artificial intelligence tools accelerate these scams by simplifying platform development and victim engagement. This reflects a shift of mature fraud ecosystems toward new financial technologies without needing entirely new criminal skills.
CVE-2026-103288 is an authorization bypass vulnerability in the Ghost blogging platform's comment like feature. It allows authenticated members to delete comment likes or dislikes they are not authorized to remove. This issue affects Ghost versions from 5.9.0 up to but not including 6.44.1. The vulnerability is fixed in version 6.44.1.
Threat actors are conducting phishing campaigns targeting Google Cloud Platform (GCP) environments by using sponsored search ads to steal privileged user credentials. They disable security controls, create service accounts with Gemini API access, and link attacker-controlled external projects to victim billing accounts. This enables unauthorized large-scale AI workload execution, causing significant unexpected charges within 24-48 hours. The attackers use evasion techniques such as geographic IP blending and demoting legitimate administrators to avoid detection and remediation.
An espionage campaign named CL-UNK-1211 targeted government and state-affiliated Linux infrastructure from October 2025 to April 2026. The attackers used AI-generated payloads with conversational annotations indicating involvement of large language models or automated AI agents. They deployed web shells, custom cross-platform backdoors including SysGoRAT malware, and open-source tools to maintain persistent access. The campaign involved extensive reconnaissance, database dumping, mailbox enumeration targeting social media accounts linked to government emails, and advanced defense evasion techniques such as disabling endpoint protections, erasing system logs, and planting deceptive red-team markers to disguise malicious activity as legitimate security assessments.
An AI-powered espionage campaign, tracked as CL-UNK-1211, targeted government and state-affiliated Linux infrastructure from October 2025 to April 2026. The attackers used AI-generated payloads, custom malware, web shells, and credential harvesting techniques to maintain persistent access and conduct reconnaissance. They deployed multiple web shells, modified PHP configurations for persistence, and used advanced anti-forensics methods including disabling endpoint protections and tampering with system logs. Credential theft was achieved through hijacked .bashrc files and PHP login page modifications. The campaign aimed to dump sensitive databases and exfiltrate government mailboxes to access social media accounts. The attackers combined open-source tools with custom cross-platform backdoors and employed AI to streamline payload creation. No specific patch or fix is mentioned.
This report describes a confirmed incident where AI-driven agents exploited an exposed, unauthenticated job-submission endpoint on an Apache Tomcat server running Spring Batch. The attackers used the open-source Cairn AI agent platform to execute JavaScript commands via the Nashorn engine, extracting data in small chunks over many requests. They obtained plaintext database credentials, enabled xp_cmdshell, escalated privileges to SYSTEM using known tools, extracted registry hives, created administrator accounts, and deleted forensic artifacts. The attack was automated and machine-driven, characterized by rapid command execution and structured output formatting.
CVE-2026-90954 is a vulnerability in BusyBox's dpkg-deb archive extraction affecting the ar archive handler. It improperly flushes deferred symlinks after each archive member rather than once after the last member, allowing crafted .deb archives with multiple data.tar members to write files outside the intended extraction directory. This can lead to arbitrary file writes to arbitrary locations when extracting malicious .deb packages. The issue affects environments using BusyBox dpkg-deb, such as Fedora and some cross-platform build or container workflows. RPM-based Red Hat systems are generally not exposed as they do not normally process .deb files. Red Hat is still analyzing product impact and no official fix has been announced yet.
Reflected Cross-Site Scripting (XSS) on the BeeTienda e-commerce platform, specifically in the latest demo version. The incident occurs due to a lack of proper sanitization of user input data in the 'search' parameter of the product list endpoint. When malicious payloads are transmitted via the 'search' parameter, they are displayed insecurely in the HTML response, allowing for the arbitrary execution of JavaScript code in the victim's browser.