| CVE-2026-2845 | An issue has been discovered in GitLab CE/EE affecting all versions from 11.2 before 18.7.5, 18.8 before 18.8.5, and 18.9 before 18.9.1 that could have allowed an authenticated user to cause denial of service by exploiting a Bitbucket Server import endpoint via repeatedly sending large responses. | medium | 2026-02-28 |
| CVE-2026-27015 | FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to version 3.23.0, a missing bounds check in `smartcard_unpack_read_size_align()` (`libfreerdp/utils/smartcard_pack.c:1703`) allows a malicious RDP server to crash the FreeRDP client via a reachable `WINPR_ASSERT` → `abort()`. The crash occurs in upstream builds where `WITH_VERBOSE_WINPR_ASSERT=ON` (default in FreeRDP 3.22.0 / current WinPR CMake defaults). Smartcard redirection must be explicitly enabled by the user (e.g., `xfreerdp /smartcard`; `/smartcard-logon` implies `/smartcard`). Version 3.23.0 fixes the issue. | medium | 2026-02-27 |
| CVE-2026-26965 | FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to version 3.23.0, in the RLE planar decode path, `planar_decompress_plane_rle()` writes into `pDstData` at `((nYDst+y) * nDstStep) + (4*nXDst) + nChannel` without verifying that `(nYDst+nSrcHeight)` fits in the destination height or that `(nXDst+nSrcWidth)` fits in the destination stride. When `TempFormat != DstFormat`, `pDstData` becomes `planar->pTempData` (sized for the desktop), while `nYDst` is only validated against the **surface** by `is_within_surface()`. A malicious RDP server can exploit this to perform a heap out-of-bounds write with attacker-controlled offset and pixel data on any connecting FreeRDP client. The OOB write reaches up to 132,096 bytes past the temp buffer end, and on the brk heap (desktop ≤ 128×128), an adjacent `NSC_CONTEXT` struct's `decode` function pointer is overwritten with attacker-controlled pixel data — control-flow–relevant corruption (function pointer overwritten) demonstrated under deterministic heap layout (`nsc->decode = 0xFF414141FF414141`). Version 3.23.0 fixes the vulnerability. | high | 2026-02-27 |
| CVE-2026-26955 | FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to version 3.23.0, a malicious RDP server can trigger a heap buffer overflow in FreeRDP clients using the GDI surface pipeline (e.g., `xfreerdp`) by sending an RDPGFX ClearCodec surface command with an out-of-bounds destination rectangle. The `gdi_SurfaceCommand_ClearCodec()` handler does not call `is_within_surface()` to validate the command rectangle against the destination surface dimensions, allowing attacker-controlled `cmd->left`/`cmd->top` (and subcodec rectangle offsets) to reach image copy routines that write into `surface->data` without bounds enforcement. The OOB write corrupts an adjacent `gdiGfxSurface` struct's `codecs*` pointer with attacker-controlled pixel data, and corruption of `codecs*` is sufficient to reach an indirect function pointer call (`NSC_CONTEXT.decode` at `nsc.c:500`) on a subsequent codec command — full instruction pointer (RIP) control demonstrated in exploitability harness. Users should upgrade to version 3.23.0 to receive a patch. | high | 2026-02-27 |
| CVE-2026-26271 | FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to version 3.23.0, a buffer overread in `freerdp_image_copy_from_icon_data()` (libfreerdp/codec/color.c) can be triggered by crafted RDP Window Icon (TS_ICON_INFO) data. The bug is reachable over the network when a client processes icon data from an RDP server (or from a man-in-the-middle). Version 3.23.0 fixes the issue. | medium | 2026-02-27 |
| CVE-2026-25997 | FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to version 3.23.0, `xf_clipboard_format_equal` reads freed `lastSentFormats` memory because `xf_clipboard_formats_free` (called from the cliprdr channel thread during auto-reconnect) frees the array while the X11 event thread concurrently iterates it in `xf_clipboard_changed`, triggering a heap use after free. Version 3.23.0 fixes the issue. | medium | 2026-02-27 |
| CVE-2026-25959 | FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to version 3.23.0, `xf_cliprdr_provide_data_` passes freed `pDstData` to `XChangeProperty` because the cliprdr channel thread calls `xf_cliprdr_server_format_data_response` which converts and uses the clipboard data without holding any lock, while the X11 event thread concurrently calls `xf_cliprdr_clear_cached_data` → `HashTable_Clear` which frees the same data via `xf_cached_data_free`, triggering a heap use after free. Version 3.23.0 fixes the issue. | medium | 2026-02-27 |
| CVE-2026-25955 | FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to version 3.23.0, `xf_AppUpdateWindowFromSurface` reuses a cached `XImage` whose `data` pointer references a freed RDPGFX surface buffer, because `gdi_DeleteSurface` frees `surface->data` without invalidating the `appWindow->image` that aliases it. Version 3.23.0 fixes the issue. | medium | 2026-02-27 |
| CVE-2026-25954 | FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to version 3.23.0, `xf_rail_server_local_move_size` dereferences a freed `xfAppWindow` pointer because `xf_rail_get_window` returns an unprotected pointer from the `railWindows` hash table, and the main thread can concurrently delete the window (via a window delete order) while the RAIL channel thread is still using the pointer. Version 3.23.0 fixes the issue. | medium | 2026-02-27 |
| CVE-2026-25953 | FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to version 3.23.0, `xf_AppUpdateWindowFromSurface` reads from a freed `xfAppWindow` because the RDPGFX DVC thread obtains a bare pointer via `xf_rail_get_window` without any lifetime protection, while the main thread can concurrently delete the window through a fastpath window-delete order. Version 3.23.0 fixes the issue. | medium | 2026-02-27 |
| CVE-2026-25952 | FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to version 3.23.0, `xf_SetWindowMinMaxInfo` dereferences a freed `xfAppWindow` pointer because `xf_rail_get_window` in `xf_rail_server_min_max_info` returns an unprotected pointer from the `railWindows` hash table, and the main thread can concurrently delete the window (via a window delete order) while the RAIL channel thread is still using the pointer. Version 3.23.0 fixes the issue. | medium | 2026-02-27 |
| CVE-2026-25942 | FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to version 3.23.0, `xf_rail_server_execute_result` indexes the global `error_code_names[]` array (7 elements, indices 0–6) with an unchecked `execResult->execResult` value received from the server, allowing an out-of-bounds read when the server sends an `execResult` value of 7 or greater. Version 3.23.0 fixes the issue. | medium | 2026-02-27 |
| CVE-2026-22721 | VMware Aria Operations contains a privilege escalation vulnerability. A malicious actor with privileges in vCenter to access Aria Operations may leverage this vulnerability to obtain administrative access in VMware Aria Operations. To remediate CVE-2026-22721, apply the patches listed in the 'Fixed Version' column of the 'Response Matrix' found in VMSA-2026-0001 https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/36947 . | high | 2026-03-04 |
| CVE-2026-1747 | GitLab has remediated an issue in GitLab EE affecting all versions from 17.11 before 18.7.5, 18.8 before 18.8.5, and 18.9 before 18.9.1 that, under certain conditions, could have allowed Developer-role users with insufficient privileges to make unauthorized modifications to protected Conan packages. | medium | 2026-02-28 |
| CVE-2026-1725 | GitLab has remediated an issue in GitLab CE/EE affecting versions from 18.9 before 18.9.1 that could have under certain conditions, allowed an unauthenticated user to cause denial of service by sending specially crafted requests to a CI jobs API endpoint. | high | 2026-02-28 |
| CVE-2026-1662 | GitLab has remediated an issue in GitLab CE/EE affecting all versions from 14.4 before 18.7.5, 18.8 before 18.8.5, and 18.9 before 18.9.1 that could have allowed an unauthenticated user to cause Denial of Service by sending specially crafted requests to the Jira events endpoint. | high | 2026-02-28 |
| CVE-2026-1388 | GitLab has remediated an issue in GitLab CE/EE affecting all versions from 9.2 before 18.7.5, 18.8 before 18.8.5, and 18.9 before 18.9.1 that could have allowed an unauthenticated user to cause regular expression denial of service by sending specially crafted input to a merge request endpoint under certain conditions. | high | 2026-02-28 |
| CVE-2026-0752 | GitLab has remediated an issue in GitLab CE/EE affecting all versions from 16.2 before 18.7.5, 18.8 before 18.8.5, and 18.9 before 18.9.1 that under certain circumstances, could have allowed an unauthenticated user to inject arbitrary scripts into the Mermaid sandbox UI. | medium | 2026-02-28 |
| CVE-2026-0542 | ServiceNow has addressed a remote code execution vulnerability that was identified in the ServiceNow AI platform. This vulnerability could enable an unauthenticated user, in certain circumstances, to execute code within the ServiceNow Sandbox. ServiceNow addressed this vulnerability by deploying a security update to hosted instances. Relevant security updates also have been provided to ServiceNow self-hosted customers and partners. Further, the vulnerability is addressed in the listed patches and hot fixes. While we are not currently aware of exploitation against customer instances, we recommend customers promptly apply appropriate updates or upgrade if they have not already done so. | critical | 2026-02-27 |
| CVE-2025-14511 | GitLab has remediated an issue in GitLab CE/EE affecting all versions from 12.2 before 18.7.5, 18.8 before 18.8.5, and 18.9 before 18.9.1 that could have allowed an unauthenticated user to cause denial of service by sending specially crafted files to the container registry event endpoint under certain conditions. | high | 2026-02-28 |
| CVE-2026-2636 | This vulnerability is caused by a CWE‑159: "Improper Handling of Invalid Use of Special Elements" weakness, which leads to an unrecoverable inconsistency in the CLFS.sys driver. This condition forces a call to the KeBugCheckEx function, allowing an unprivileged user to trigger a system crash. Microsoft silently fixed this vulnerability in the September 2025 cumulative update for Windows 11 2024 LTSC and Windows Server 2025. Windows 25H2 (released in September) was released with the patch. Windows 1123h2 and earlier versions remain vulnerable. | medium | 2026-02-27 |
| CVE-2026-25941 | FreeRDP is a free implementation of the Remote Desktop Protocol. Versions on the 2.x branch prior to to 2.11.8 and on the 3.x branch prior to 3.23.0 have an out-of-bounds read vulnerability in the FreeRDP client's RDPGFX channel that allows a malicious RDP server to read uninitialized heap memory by sending a crafted WIRE_TO_SURFACE_2 PDU with a `bitmapDataLength` value larger than the actual data in the packet. This can lead to information disclosure or client crashes when a user connects to a malicious server. Versions 2.11.8 and 3.23.0 fix the issue. | high | 2026-02-27 |
| CVE-2026-25736 | Rucio is a software framework that provides functionality to organize, manage, and access large volumes of scientific data using customizable policies. Versions prior to 35.8.3, 38.5.4, and 39.3.1 have a stored Cross-Site Scripting (XSS) vulnerability in the Custom RSE Attribute of the WebUI where attacker-controlled input is persisted by the backend and later rendered in the WebUI without proper output encoding. This allows arbitrary JavaScript execution in the context of the WebUI for users who view affected pages, potentially enabling session token theft or unauthorized actions. Versions 35.8.3, 38.5.4, and 39.3.1 fix the issue. | medium | 2026-02-27 |
| CVE-2026-25735 | Rucio is a software framework that provides functionality to organize, manage, and access large volumes of scientific data using customizable policies. Versions prior to 35.8.3, 38.5.4, and 39.3.1 have a stored Cross-Site Scripting (XSS) vulnerability in the Identity Name of the WebUI where attacker-controlled input is persisted by the backend and later rendered in the WebUI without proper output encoding. This allows arbitrary JavaScript execution in the context of the WebUI for users who view affected pages, potentially enabling session token theft or unauthorized actions. Versions 35.8.3, 38.5.4, and 39.3.1 fix the issue. | medium | 2026-02-27 |
| CVE-2026-25734 | Rucio is a software framework that provides functionality to organize, manage, and access large volumes of scientific data using customizable policies. Versions prior to 35.8.3, 38.5.4, and 39.3.1 have a stored Cross-Site Scripting (XSS) vulnerability in the RSE metadata of the WebUI where attacker-controlled input is persisted by the backend and later rendered in the WebUI without proper output encoding. This allows arbitrary JavaScript execution in the context of the WebUI for users who view affected pages, potentially enabling session token theft or unauthorized actions. Versions 35.8.3, 38.5.4, and 39.3.1 fix the issue. | medium | 2026-02-27 |
| CVE-2026-25733 | Rucio is a software framework that provides functionality to organize, manage, and access large volumes of scientific data using customizable policies. Versions prior to 35.8.3, 38.5.4, and 39.3.1 have a stored Cross-Site Scripting (XSS) vulnerability in the Custom Rules function of the WebUI where attacker-controlled input is persisted by the backend and later rendered in the WebUI without proper output encoding. This allows arbitrary JavaScript execution in the context of the WebUI for users who view affected pages, potentially enabling session token theft or unauthorized actions. Versions 35.8.3, 38.5.4, and 39.3.1 fix the issue. | medium | 2026-02-27 |
| CVE-2026-25138 | Rucio is a software framework that provides functionality to organize, manage, and access large volumes of scientific data using customizable policies. Prior to versions 35.8.3, 38.5.4, and 39.3.1, the WebUI login endpoint returns distinct error messages depending on whether a supplied username exists, allowing unauthenticated attackers to enumerate valid usernames. Versions 35.8.3, 38.5.4, and 39.3.1 fix the issue. | medium | 2026-02-27 |
| CVE-2026-25136 | Rucio is a software framework that provides functionality to organize, manage, and access large volumes of scientific data using customizable policies. A reflected Cross-site Scripting vulnerability was located in versions prior to 35.8.3, 38.5.4, and 39.3.1 in the rendering of the ExceptionMessage of the WebUI 500 error which could allow attackers to steal login session tokens of users who navigate to a specially crafted URL. Versions 35.8.3, 38.5.4, and 39.3.1 fix the issue. | medium | 2026-02-27 |
| CVE-2026-22720 | VMware Aria Operations contains a stored cross-site scripting vulnerability. A malicious actor with privileges to create custom benchmarks may be able to inject script to perform administrative actions in VMware Aria Operations. To remediate CVE-2026-22720, apply the patches listed in the 'Fixed Version' column of the 'Response Matrix' of VMSA-2026-0001 https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/36947https:// . | critical | 2026-03-04 |
| CVE-2026-22719 | VMware Aria Operations contains a command injection vulnerability. A malicious unauthenticated actor may exploit this issue to execute arbitrary commands which may lead to remote code execution in VMware Aria Operations while support-assisted product migration is in progress. To remediate CVE-2026-22719, apply the patches listed in the 'Fixed Version' column of the ' Response Matrix https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/36947 ' in VMSA-2026-0001 Workarounds for CVE-2026-22719 are documented in the 'Workarounds' column of the ' Response Matrix https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/36947 ' in VMSA-2026-0001 | high | 2026-03-04 |
| CVE-2025-3525 | GitLab has remediated an issue in GitLab CE/EE affecting all versions from 9.0 before 18.7.5, 18.8 before 18.8.5, and 18.9 before 18.9.1 that could have, under certain circumstances, allowed an authenticated user with certain access to cause Denial of Service by creating specially crafted CI triggers via the API. | medium | 2026-02-27 |
| CVE-2025-14103 | GitLab has remediated an issue in GitLab CE/EE affecting all versions from 17.7 before 18.7.5, 18.8 before 18.8.5, and 18.9 before 18.9.1 that could have allowed an unauthorized user with Developer-role permissions to set pipeline variables for manually triggered jobs under certain conditions. | medium | 2026-02-27 |
| CVE-2026-3221 | Sensitive user account information is not encrypted in the database in Devolutions Server 2025.3.14 and earlier, which allows an attacker with access to the database to obtain sensitive user information via direct database access. | medium | 2026-02-28 |
| CVE-2026-25930 | OpenEMR is a free and open source electronic health records and medical practice management application. Prior to version 8.0.0, the Layout-Based Form (LBF) printable view accepts `formid` and `visitid` (or `patientid`) from the request and does not verify that the form belongs to the current user’s authorized patient/encounter. An authenticated user with LBF access can enumerate form IDs and view or print any patient’s encounter forms. Version 8.0.0 fixes the issue. | medium | 2026-02-27 |
| CVE-2026-25929 | OpenEMR is a free and open source electronic health records and medical practice management application. Prior to version 8.0.0, the document controller’s `patient_picture` context serves the patient’s photo by document ID or patient ID without verifying that the current user is authorized to access that patient. An authenticated user with document ACL can supply another patient’s ID and retrieve their photo. Version 8.0.0 fixes the issue. | medium | 2026-02-27 |
| CVE-2026-25927 | OpenEMR is a free and open source electronic health records and medical practice management application. Prior to version 8.0.0, the DICOM viewer state API (e.g. upload or state save/load) accepts a document ID (`doc_id`) without verifying that the document belongs to the current user’s authorized patient or encounter. An authenticated user can read or modify DICOM viewer state (e.g. annotations, view settings) for any document by enumerating document IDs. Version 8.0.0 fixes the issue. | high | 2026-02-27 |
| CVE-2026-25746 | OpenEMR is a free and open source electronic health records and medical practice management application. Versions prior to 8.0.0 contain a SQL injection vulnerability in prescription that can be exploited by authenticated attackers. The vulnerability exists due to insufficient input validation in the prescription listing functionality. Version 8.0.0 fixes the vulnerability. | high | 2026-02-27 |
| CVE-2026-25743 | OpenEMR is a free and open source electronic health records and medical practice management application. Prior to version 8.0.0, users with the "Forms administration" role can fill questionnaires ("forms") in patient encounters. The answers to the forms are displayed on the encounter page and in the visit history for the users with the same role. There exists a stored cross-site scripting (XSS) vulnerability in the function to display the form answers, allowing any authenticated attacker with the specific role to insert arbitrary JavaScript into the system by entering malicious payloads to the form answers. The JavaScript code is later executed by any user with the form role when viewing the form answers in the patient encounter pages or visit history. Version 8.0.0 fixes the issue. | high | 2026-02-27 |
| CVE-2026-25476 | OpenEMR is a free and open source electronic health records and medical practice management application. Prior to version 8.0.0, the session expiration check in `library/auth.inc.php` runs only when `skip_timeout_reset` is not present in the request. When `skip_timeout_reset=1` is sent, the entire block that calls `SessionTracker::isSessionExpired()` and forces logout on timeout is skipped. As a result, any request that includes this parameter (e.g. from auto-refresh pages like the Patient Flow Board) never runs the expiration check: expired sessions can continue to access data indefinitely, abandoned workstations stay active, and an attacker with a stolen session cookie can keep sending `skip_timeout_reset=1` to avoid being logged out. Version 8.0.0 fixes the issue. | high | 2026-02-28 |
| CVE-2026-25220 | OpenEMR is a free and open source electronic health records and medical practice management application. Prior to version 8.0.0, the Message Center accepts the URL parameter `show_all=yes` and passes it to `getPnotesByUser()`, which returns all internal messages (all users’ notes). The backend does not verify that the requesting user is an administrator before honoring `show_all=yes`. The "Show All" link is also visible to non-admin users. As a result, any authenticated user can view the entire internal message list by requesting `messages.php?show_all=yes`. Version 8.0.0 patches the issue. | high | 2026-02-27 |
| CVE-2026-25164 | OpenEMR is a free and open source electronic health records and medical practice management application. Prior to version 8.0.0, the REST API route table in `apis/routes/_rest_routes_standard.inc.php` does not call `RestConfig::request_authorization_check()` for the document and insurance routes. Other patient routes in the same file (e.g. encounters, patients/med) call it with the appropriate ACL. As a result, any valid API bearer token can access or modify every patient's documents and insurance data, regardless of the token’s OpenEMR ACLs—effectively exposing all document and insurance PHI to any authenticated API client. Version 8.0.0 patches the issue. | high | 2026-02-27 |
| CVE-2026-24908 | OpenEMR is a free and open source electronic health records and medical practice management application. Prior to version 8.0.0, an SQL injection vulnerability in the Patient REST API endpoint allows authenticated users with API access to execute arbitrary SQL queries through the `_sort` parameter. This could potentially lead to database access, PHI (Protected Health Information) exposure, and credential compromise. The issue occurs when user-supplied sort field names are used in ORDER BY clauses without proper validation or identifier escaping. Version 8.0.0 fixes the issue. | medium | 2026-02-27 |
| CVE-2026-24890 | OpenEMR is a free and open source electronic health records and medical practice management application. Prior to version 8.0.0, an authorization bypass vulnerability in the patient portal signature endpoint allows authenticated portal users to upload and overwrite provider signatures by setting `type=admin-signature` and specifying any provider user ID. This could potentially lead to signature forgery on medical documents, legal compliance violations, and fraud. The issue occurs when portal users are allowed to modify provider signatures without proper authorization checks. Version 8.0.0 fixes the issue. | medium | 2026-02-27 |
| CVE-2026-24487 | OpenEMR is a free and open source electronic health records and medical practice management application. Prior to version 8.0.0, an authorization bypass vulnerability in the FHIR CareTeam resource endpoint allows patient-scoped FHIR tokens to access care team data for all patients instead of being restricted to only the authenticated patient's data. This could potentially lead to unauthorized disclosure of Protected Health Information (PHI), including patient-provider relationships and care team structures across the entire system. The issue occurs because the `FhirCareTeamService` does not implement the `IPatientCompartmentResourceService` interface and does not pass the patient binding parameter to the underlying service, bypassing the patient compartment filtering mechanism. Version 8.0.0 contains a patch for this issue. | high | 2026-02-27 |
| CVE-2026-24005 | Kruise provides automated management of large-scale applications on Kubernetes. Prior to versions 1.8.3 and 1.7.5, PodProbeMarker allows defining custom probes with TCPSocket or HTTPGet handlers. The webhook validation does not restrict the Host field in these probe configurations. Since kruise-daemon runs with hostNetwork=true, it executes probes from the node network namespace. An attacker with PodProbeMarker creation permission can specify arbitrary Host values to trigger SSRF from the node, perform port scanning, and receive response feedback through NodePodProbe status messages. Versions 1.8.3 and 1.7.5 patch the issue. | high | 2026-03-05 |
| CVE-2026-23627 | OpenEMR is a free and open source electronic health records and medical practice management application. Prior to version 8.0.0, an SQL injection vulnerability in the Immunization module allows any authenticated user to execute arbitrary SQL queries, leading to complete database compromise, PHI exfiltration, credential theft, and potential remote code execution. The vulnerability exists because user-supplied `patient_id` values are directly concatenated into SQL WHERE clauses without parameterization or escaping. Version 8.0.0 patches the issue. | high | 2026-02-27 |
| CVE-2026-3194 | A flaw has been found in Chia Blockchain 2.1.0. The affected element is the function send_transaction/get_private_key of the component RPC Server Master Passphrase Handler. This manipulation causes missing authentication. The attack can only be executed locally. The attack's complexity is rated as high. The exploitability is described as difficult. The exploit has been published and may be used. The vendor was informed early via email. A separate report via bugbounty was rejected with the reason "This is by design. The user is responsible for host security". | low | 2026-03-05 |
| CVE-2026-27850 | Due to an improperly configured firewall rule, the router will accept any connection on the WAN port with the source port 5222, exposing all services which are normally only accessible through the local network. This issue affects MR9600: 1.0.4.205530; MX4200: 1.0.13.210200. | high | 2026-02-27 |
| CVE-2026-27795 | LangChain is a framework for building LLM-powered applications. Prior to version 1.1.8, a redirect-based Server-Side Request Forgery (SSRF) bypass exists in `RecursiveUrlLoader` in `@langchain/community`. The loader validates the initial URL but allows the underlying fetch to follow redirects automatically, which permits a transition from a safe public URL to an internal or metadata endpoint without revalidation. This is a bypass of the SSRF protections introduced in 1.1.14 (CVE-2026-26019). Users should upgrade to `@langchain/community` 1.1.18, which validates every redirect hop by disabling automatic redirects and re-validating `Location` targets before following them. In this version, automatic redirects are disabled (`redirect: "manual"`), each 3xx `Location` is resolved and validated with `validateSafeUrl()` before the next request, and a maximum redirect limit prevents infinite loops. | medium | 2026-02-27 |
| CVE-2026-27794 | LangGraph Checkpoint defines the base interface for LangGraph checkpointers. Prior to version 4.0.0, a Remote Code Execution vulnerability exists in LangGraph's caching layer when applications enable cache backends that inherit from `BaseCache` and opt nodes into caching via `CachePolicy`. Prior to `langgraph-checkpoint` 4.0.0, `BaseCache` defaults to `JsonPlusSerializer(pickle_fallback=True)`. When msgpack serialization fails, cached values can be deserialized via `pickle.loads(...)`. Caching is not enabled by default. Applications are affected only when the application explicitly enables a cache backend (for example by passing `cache=...` to `StateGraph.compile(...)` or otherwise configuring a `BaseCache` implementation), one or more nodes opt into caching via `CachePolicy`, and the attacker can write to the cache backend (for example a network-accessible Redis instance with weak/no auth, shared cache infrastructure reachable by other tenants/services, or a writable SQLite cache file). An attacker must be able to write attacker-controlled bytes into the cache backend such that the LangGraph process later reads and deserializes them. This typically requires write access to a networked cache (for example a network-accessible Redis instance with weak/no auth or shared cache infrastructure reachable by other tenants/services) or write access to local cache storage (for example a writable SQLite cache file via permissive file permissions or a shared writable volume). Because exploitation requires write access to the cache storage layer, this is a post-compromise / post-access escalation vector. LangGraph Checkpoint 4.0.0 patches the issue. | medium | 2026-02-27 |
