Vulnerabilities
All
My stack
Known exploited
Public exploit
Critical (9+)
Likely exploited (EPSS ≥ 50%)
Changed, 24 h
New, 24 h
Sort:
risk
modified
published
CVSS
EPSS
| CVE | CVSS | EPSS | Vendor, product | Description | Modified |
|---|---|---|---|---|---|
| CVE-2026-48520 | 6.1 medium | 0.00% | langflow | Langflow is a tool for building and deploying AI-powered agents and workflows. Prior to 1.10.0, the "Shareable Playground" (or "Public Flows" in code) contains a potential arbitrary file-read vulnerability, depending on the exact flow configuration used. By making a flow public, public execution of the flow is allowed. The execution request can contain a list of files that gets read by Langflow and fed into the LLM. The files path can be any path supported by the storage - it can be either a local file or S3 path if supported by the local configuration This vulnerability is fixed in 1.10.0. | Mon 26 Oct |
| CVE-2026-40979 | 6.1 medium | 0.00% | vmware, spring_ai | In Spring AI, having access to a shared environment can expose the ONNX model used by the application. Affected versions: Spring AI: 1.0.0 - 1.0.5 (fixed in 1.0.6), 1.1.0 - 1.1.4 (fixed in 1.1.5) | 26 Oct |
| CVE-2026-55423 | 6.1 medium | 0.00% | langflow | Langflow is a tool for building and deploying AI-powered agents and workflows. Prior to 1.7.0, the logout button does not clear the session. The previous user stays logged in unless another user explicitly logs in. This vulnerability is fixed in 1.7.0. | 25 Oct |
| CVE-2026-54714 | 6.1 medium | 0.00% | no CPE yet | Logto is the modern, open-source auth infrastructure for SaaS and AI apps. Prior to 1.41.0, @logto/core reflected the SAML RelayState, SAMLResponse, and actionUrl into a Logto-origin auto-submit HTML form in packages/core/src/saml-application/SamlApplication/utils.ts without HTML-attribute escaping. A SAML application flow with a crafted RelayState from GET or POST /api/saml/:id/authn could inject script that runs on the Logto tenant origin after the user completes login. This issue is fixed in version 1.41.0. | 25 Oct |
| CVE-2026-13234 | 6.1 medium | 0.00% | artificial_intelligence_project, artificial_intelligence | Improper Neutralization of Input During Web Page Generation ("Cross-site Scripting") vulnerability in Drupal AI (Artificial Intelligence) allows Cross-Site Scripting (XSS). This issue affects AI (Artificial Intelligence) versions: from 0.0.0 to 1.2.17, from 1.3.0 to 1.3.8, from 1.4.0 to 1.4.3. | 25 Oct |
| CVE-2026-46341 | 6.1 medium | 0.00% | no CPE yet | The Apify MCP server enables AI agents to extract data from websites using ready-made scrapers, crawlers, and automation tools available on the Apify Store. Prior to 0.9.21, the fetch-apify-docs tool in src/tools/common/fetch_apify_docs.ts validates allowlisted documentation domains with String.startsWith() rather than URL hostname comparison, allowing attacker-controlled URLs such as `https://docs.apify.com.evil.com/` and `https://[email protected]/` to pass the ALLOWED_DOC_DOMAINS check and return arbitrary fetched content to the LLM. This issue is fixed in version 0.9.21. | 24 Oct |
| CVE-2026-58413 | 6.1 medium | 0.00% | network-ai | Network-AI is a TypeScript/Node.js multi-agent orchestrator. Prior to version 5.12.2, `EnvironmentManager.restore(env, backupId)` computes the backup path with `join(envDir, '.backups', backupId)` and only checks that this path exists. It does not resolve the result or verify that it remains under `data/<env>/.backups`. A caller can pass a traversal backup ID such as `../../../outside/source-dir` to restore files from an arbitrary directory into the target environment data directory. The issue is fixed in v5.12.2. `restore()` now validates `backupId` against `/^[\w\-]+$/` and asserts `dirname(resolve(join(backupsDir, backupId))) === resolve(backupsDir)` before touching the filesystem. Backup IDs containing path separators or `..` are rejected, so a crafted ID can no longer copy directories from outside `.backups/` into the environment. | 24 Oct |
| CVE-2026-47128 | 6.1 medium | 0.00% | no CPE yet | nono is software that allows users to run AI agents in a zero-latency sandbox. Prior to version 0.55.0, the nono Landlock/seccomp policies allow access to local Unix domain sockets (concrete and abstract). This allows an easy sandbox escape by talking to the per-user systemd dbus socket. Version 0.55.0 patches the issue. | 24 Oct |
| CVE-2025-9952 | 6.1 medium | 0.00% | no CPE yet | The Trinity Audio – Text to Speech AI audio player to convert content into audio plugin for WordPress is vulnerable to Reflected Cross-Site Scripting via the 'range-date' parameter in all versions up to, and including, 5.20.2 due to insufficient input sanitization and output escaping. This makes it possible for unauthenticated attackers to inject arbitrary web scripts in pages that execute if they can successfully trick a user into performing an action such as clicking on a link. | 11 Oct |
| CVE-2026-65698 | 6.0 medium | 0.00% | no CPE yet | Void through 1.3.4 contains a path traversal vulnerability in the AI agent file-reading tools that allows network-adjacent attackers to read arbitrary host files outside the open workspace by injecting instructions into content the agent processes. Attackers can supply absolute paths or file:// URIs through the read_file, ls_dir, get_dir_tree, and search_* tools, which lack workspace confinement and bypass the approval gate, enabling silent exfiltration of sensitive files such as SSH private keys or cloud credentials via subsequent tool calls. | 24 Oct |
| CVE-2026-110611 | 6.0 medium | 0.00% | no CPE yet | REBUILD through 4.5.0 contains an authorization bypass vulnerability in AiBot2Controller /aibot2 endpoints that act on a caller-supplied chatid without verifying ownership. Authenticated low-privilege users can supply another user's chatid to chat-init, chat-delete, chat-feedback, or chat-stream-stop to read, delete, exfiltrate, or interrupt AI assistant conversations. | 20 Oct |
| CVE-2026-110848 | 6.0 medium | 0.00% | no CPE yet | n8n before 1.123.83, 2.x before 2.41.4, and 2.42.x before 2.42.1 contains a prototype pollution vulnerability in the AI workflow builder, which merges workflow connections using node names as object members without checking for reserved keys. Authenticated users can create nodes named __proto__ through the API and invoke the connect-nodes operation to pollute the global object prototype, disrupting the instance until restart. | 16 Oct |
| CVE-2026-40966 | 5.9 medium | 0.00% | vmware, spring_ai | In Spring AI, an attacker can bypass conversation isolation and exfiltrate sensitive memory from other users’ chat histories, including secrets and credentials, by injecting filter logic through conversationId. Only applications that use VectorStoreChatMemoryAdvisor and pass user-supplied input as a conversationId are affected. | Mon 26 Oct |
| CVE-2026-54323 | 5.9 medium | 0.00% | no CPE yet | Daytona is a secure and elastic infrastructure runtime for AI-generated code execution and agent workflows. Prior to 0.185.0, the daemon's git clone implementation disabled TLS certificate verification. When a clone request carried Git credentials, the daemon sent the HTTP Basic Authorization header to the remote over a connection whose certificate was never validated, on both the go-git and native git CLI code paths. An attacker able to intercept clone traffic could present any TLS certificate, capture the Git credentials supplied for the clone, and serve tampered repository content into the sandbox. This vulnerability is fixed in 0.185.0. | 25 Oct |
| CVE-2026-61643 | 5.9 medium | 0.00% | no CPE yet | FastGPT is a knowledge-based AI application platform. From 4.14.17 until 4.15.0-beta5, an authenticated FastGPT user can save a workflow node that points to another user's private HTTP toolset by using a crafted saved tool id such as http-<victim_toolset_app_id>/<tool_name>. The normal toolset routes deny access, but the workflow save and runtime path did not apply the same authorization check to the referenced toolset, allowing /api/v2/chat/completions to resolve the saved reference and execute the victim-owned HTTP tool. This issue is fixed in version 4.15.0-beta5. | 24 Oct |
| CVE-2026-58482 | 5.9 medium | 0.00% | network-ai | Network-AI, a TypeScript/Node.js multi-agent orchestrator, has a shipped, exported, documented feature called `ApprovalInbox` (`lib/approval-inbox.ts`). It is the network surface of the human-in-the-loop Approval Gate, which `ApprovalGate` uses to require explicit human approval for high-risk operations. The HTTP server it exposes has no authentication of any kind and sets `Access-Control-Allow-Origin: *` on every route, including the state-changing `POST /approvals/:id/approve` and `/deny`. As a result, in versions 5.0.0 through 5.12.1, any party who can send an HTTP request to the inbox port — a co-located process, a container/SSRF on the same host, a remote client when the operator binds a non-loopback address, or any website the operator visits in a browser (via the wildcard CORS) — can enumerate pending approvals and approve them, defeating the entire human-in-the-loop control and causing the gated high-risk action (e.g. a shell command the agent was holding for review) to execute without consent. This issue is fixed in v5.12.2. `ApprovalInbox` now accepts a `secret` option. When set, the mutating endpoints `POST /:id/approve` and `POST /:id/deny` require an `Authorization: Bearer <secret>` header, validated in constant time with `crypto.timingSafeEqual`. `startServer()` already binds to `127.0.0.1` by default; operators exposing the inbox on a network must set a secret. | 24 Oct |
| CVE-2025-61589 | 5.9 medium | 0.00% | anysphere, cursor | Cursor is a code editor built for programming with AI. In versions 1.6 and below, Mermaid (a to render diagrams) allows embedding images which then get rendered by Cursor in the chat box. An attacker can use this to exfiltrate sensitive information to a third-party attacker controlled server through an image fetch after successfully performing a prompt injection. A malicious model (or hallucination/backdoor) might also trigger this exploit at will. This issue requires prompt injection from malicious data (web, image upload, source code) in order to exploit. In that case, it can send sensitive information to an attacker-controlled external server. Some additional bypasses not covered in the initial fix to this issue were discovered, see GHSA-43wj-mwcc-x93p. This issue is fixed in version 1.7. | 19 Oct |
| CVE-2026-108591 | 5.9 medium | 0.00% | no CPE yet | InnoShop 0.9.2 contains a local file disclosure vulnerability that allows authenticated administrators with files_create permission to read server files by abusing the AI Core MCP file_upload tool's source argument. Attackers can supply file:// or php:// stream wrappers passed to file_get_contents(), storing contents on the public media disk to expose the .env file with APP_KEY and database credentials. | 18 Oct |
| CVE-2026-55946 | 5.9 medium | 0.00% | microsoft, copilot | Ai command injection in Microsoft Copilot allows an unauthorized attacker to disclose information over a network. | 14 Oct |
| CVE-2025-60160 | 5.9 medium | 0.00% | no CPE yet | Improper Neutralization of Input During Web Page Generation ('Cross-site Scripting') vulnerability in sharkthemes Smart Related Products ai-related-products allows Stored XSS.This issue affects Smart Related Products: from n/a through <= 2.0.8. | 9 Oct |
| CVE-2026-62314 | 5.8 medium | 0.00% | no CPE yet | Anubis is a Web AI Firewall Utility that challenges users' connections in order to protect upstream resources from scraper bots. From 1.22.0 until 1.26.0-pre1, lib/policy/checker.go PathChecker.Check() trusted the client-controlled X-Original-URI header before matching r.URL.Path, allowing an HTTP client to match default data/common/keep-internet-working.yaml ALLOW rules such as ^/\.well-known/.*$ and bypass the Anubis challenge. This issue is fixed in version 1.26.0-pre1. | 24 Oct |
| CVE-2026-105743 | 5.8 medium | 0.00% | docling | Docling simplifies document processing by parsing diverse formats and providing integrations with the generative AI ecosystem. From 2.91.0 until 2.132.0, validate_url_safety in docling/backend/utils/image_resource_loader.py validates a hostname with a single IPv4 lookup and then allows the HTTP client to resolve and parse the original URL again, permitting DNS rebinding, mixed public and internal address records, and backslash authority parser disagreement to reach internal services. HTMLBackendOptions(render_page=True) also allows HTTP and HTTPS browser requests without validating their resolved destination. Exploitation requires remote fetching to be enabled, and response content is exposed only when it is decoded as an image or passively rendered in a page screenshot. This issue is fixed in 2.132.0. | 8 Oct |
| CVE-2026-15737 | 5.7 medium | 0.00% | no CPE yet | AWS Bedrock AgentCore Python SDK is an open-source Python library that provides client tools for building AI agents on the Amazon Bedrock AgentCore platform. Unintended logging of sensitive user content in the OpenTelemetry instrumentation in AWS Bedrock AgentCore Python SDK versions 1.4.8 and 1.5.0 might allow a local authenticated user with access to CloudWatch Logs to access raw user prompts and agent responses containing sensitive data via span attributes. The SDK wrote raw user prompts and complete agent responses into OpenTelemetry span attributes on every invocation without filtering or masking. These spans flow into the customer's aws/spans CloudWatch log group, exposing sensitive content to any principal with log read access. We recommend you upgrade to version 1.5.1 or later. Users who ran affected versions should also review and purge sensitive content from their aws/spans CloudWatch log groups. | 24 Oct |
| CVE-2026-102682 | 5.6 medium | 0.00% | no CPE yet | Pipecat is an open-source Python framework for building real-time voice and multimodal conversational agents. From 0.0.98 until 1.6.0, the pipecat-ai package's built-in development runner exposes unauthenticated endpoints including /start, /daily, and /daily-dialin-webhook that can create bot sessions and Daily rooms. The runner binds to localhost by default and is intended only for local development, so exploitation requires an operator to expose it to untrusted traffic through --host 0.0.0.0, a published port, or a tunnel without access controls. An unauthenticated caller can consume local resources and third-party quotas and, depending on configuration, receive a room URL and join token. Version 1.6.0 adds a prominent development-runner banner and documentation that clarifies the unauthenticated boundary and directs production deployments to an authenticated dispatcher or managed runtime, while the endpoint behavior remains by design. This issue is fixed in version 1.6.0. | 16 Oct |
| CVE-2026-39417 | 5.5 medium | 0.00% | maxkb | MaxKB is an open-source AI assistant for enterprise. Versions 2.7.1 and below contain an incomplete fix for CVE-2025-53928, where a Remote Code Execution vulnerability still exists in the MCP node of the workflow engine. MaxKB only restricts the referencing code path (loading MCP config from the database). The else branch, responsible for loading mcp_servers directly from user-supplied JSON remains completely unpatched. Since mcp_source is an optional field (required=False), an attacker can simply omit it or set it to any non-referencing value to bypass the fix. By calling the workflow creation API directly with a crafted JSON payload, an attacker can inject a complete MCP node configuration with stdio transport, arbitrary command, and args — achieving RCE when the workflow is triggered via chat. This issue has been fixed in version 2.8.0. | Mon 26 Oct |
| CVE-2026-40608 | 5.5 medium | 0.00% | dayuanjiang, next_ai_draw.io | Next AI Draw.io is a next.js web application that integrates AI capabilities with draw.io diagrams. Prior to 0.4.15, the embedded HTTP sidecar contains three POST handlers (/api/state, /api/restore, and /api/history-svg) that process incoming requests by accumulating the entire request body into a JavaScript string without any size limitations. Node.js buffers the entire payload in the V8 heap. Sending a sufficiently large body (e.g., 500 MiB or more) will exhaust the process heap memory, leading to an Out-of-Memory (OOM) error that crashes the MCP server. This vulnerability is fixed in 0.4.15. | Mon 26 Oct |
| CVE-2026-46383 | 5.5 medium | 0.00% | no CPE yet | Microsoft APM is an open-source, community-driven dependency manager for AI agents. Prior to 0.13.0, Microsoft APM contains a Windows-specific archive extraction boundary failure in the legacy-bundle probe used by apm install <bundle> on supported Python 3.10 and 3.11 runtimes. When apm install is given a local .tar.gz that is not recognized as a plugin-format bundle, APM probes whether it is a legacy --format apm bundle. On Python versions earlier than 3.12, that probe extracts untrusted tar members with raw tar.extractall() without rejecting Windows absolute member names such as D:/.... This vulnerability is fixed in 0.13.0. | Mon 26 Oct |
| CVE-2026-45046 | 5.5 medium | 0.00% | no CPE yet | Gryph provides a security layer for AI coding agents. Prior to 0.7.0, Gryph implements logging levels that determine what content is logged to a local sqlite database. The README incorrectly mentions that the default log level is minimal while it is standard. Source code review shows sensitive file-write content remains in the stored payload as ContentPreview, OldString, or NewString at the default standard logging level and at full. This leads to logging of potentially sensitive file content in the local sqlite database, violating Gryphs sensitive file filter and log level contracts. This vulnerability is fixed in 0.7.0. | Mon 26 Oct |
| CVE-2026-44022 | 5.5 medium | 0.00% | docling | Docling simplifies document processing by parsing diverse formats and providing integrations with the generative AI ecosystem. From 2.73.0 until 2.91.0, he LaTeX backend's handling of \includegraphics, \input, and \include commands lacked path containment validation. Attackers could craft malicious LaTeX documents with path traversal sequences to read arbitrary files from the file system accessible to the process, include sensitive files in the converted document output, or potentially access configuration files, credentials, or other sensitive data This vulnerability is fixed in 2.91.0. | 25 Oct |
| CVE-2026-16016 | 5.5 medium | 0.00% | no CPE yet | A vulnerability was identified in poco-ai poco-claw up to 0.5.4. This issue affects the function run_task of the file executor/app/api/v1/task.py. The manipulation of the argument callback_url leads to server-side request forgery. The attack is possible to be carried out remotely. The exploit is publicly available and might be used. The reported GitHub issue was closed automatically due to inactivity. | 24 Oct |
| CVE-2026-58414 | 5.5 medium | 0.00% | network-ai | Network-AI is a TypeScript/Node.js multi-agent orchestrator. Prior to version 5.12.2, `EnvironmentManager.backup()` recursively collects files using `_collectBackupFiles()`. `_collectBackupFiles()` uses `statSync(full)`, which follows symlinks. If `data/<env>` contains a symlink to a directory outside the environment root, backup recursion follows the symlink and copies external files into `data/<env>/.backups/<backupId>/`. An attacker who can place a symlink under the environment data directory can cause backup operations to disclose files outside the environment root into backup artifacts. The issue is fixed in v5.12.2. `_collectBackupFiles()` now uses `lstatSync` instead of `statSync` and skips any entry where `isSymbolicLink()` is true. Symlinks are never traversed, so `backup()` can no longer follow a link out of the environment root and copy external files into a backup artifact. | 24 Oct |
| CVE-2026-17534 | 5.5 medium | 0.00% | no CPE yet | Kimi Code (@moonshot-ai/kimi-code) before 0.27.0 implements FetchURL SSRF hardening as a static hostname and IP-literal denylist in assertSafeFetchTarget, without resolving DNS or re-validating hosts after HTTP redirects. An attacker who can influence a FetchURL call (for example via prompt injection) can supply a crafted public hostname that resolves to loopback or another internal address, or a public URL that redirects to such a target, and thereby reach internal network services that the denylist was intended to block. FetchURL is included in the default auto-approve tool set, so the call does not require interactive user confirmation in manual mode. | 24 Oct |
| CVE-2026-18770 | 5.5 medium | 0.00% | no CPE yet | A vulnerability has been found in vibesurf-ai VibeSurf up to cd6e519d507cdd4d63061300bf60fb176e1f57e0. Impacted is an unknown function of the file /code of the component Python Validation Handler. The manipulation leads to code injection. Remote exploitation of the attack is possible. This product follows a rolling release approach for continuous delivery, so version details for affected or updated releases are not provided. The vendor was contacted early about this disclosure but did not respond in any way. | 23 Oct |
| CVE-2026-18647 | 5.5 medium | 0.00% | no CPE yet | A security vulnerability has been detected in jina-ai reader up to 1574bfd380d249c86c82db4dace0d9c8fe17e2b1. This issue affects the function isValidTLD of the file /backend/functions/src/cloud-functions/crawler.ts of the component Crawler/Puppeteer. The manipulation leads to server-side request forgery. Remote exploitation of the attack is possible. The exploit has been disclosed publicly and may be used. This product follows a rolling release approach for continuous delivery, so version details for affected or updated releases are not provided. The vendor was contacted early about this disclosure but did not respond in any way. | 23 Oct |
| CVE-2026-112145 | 5.5 medium | 0.00% | no CPE yet | A security vulnerability has been detected in speaches-ai speaches up to 0.9.0-rc.3. This vulnerability affects the function PyannoteDiarizationModelManager._load_fn of the file src/speaches/executors/pyannote_diarization.py of the component Pyannote Diarization Model Loading. The manipulation of the argument model leads to use of externally-controlled input to select classes or code. It is possible to initiate the attack remotely. The exploit has been disclosed publicly and may be used. The feature was introduced in PR #628 (feat/switch to pyannote audio), merged as commit 870e1e16a73452496e99d49f759eabd692a8e0e0. The project was informed of the problem early through an issue report but has not responded yet. | 22 Oct |
| CVE-2025-39684 | 5.5 medium | 0.00% | linux, debian, linux_kernel, debian_linux | In the Linux kernel, the following vulnerability has been resolved: comedi: Fix use of uninitialized memory in do_insn_ioctl() and do_insnlist_ioctl() syzbot reports a KMSAN kernel-infoleak in `do_insn_ioctl()`. A kernel buffer is allocated to hold `insn->n` samples (each of which is an `unsigned int`). For some instruction types, `insn->n` samples are copied back to user-space, unless an error code is being returned. The problem is that not all the instruction handlers that need to return data to userspace fill in the whole `insn->n` samples, so that there is an information leak. There is a similar syzbot report for `do_insnlist_ioctl()`, although it does not have a reproducer for it at the time of writing. One culprit is `insn_rw_emulate_bits()` which is used as the handler for `INSN_READ` or `INSN_WRITE` instructions for subdevices that do not have a specific handler for that instruction, but do have an `INSN_BITS` handler. For `INSN_READ` it only fills in at most 1 sample, so if `insn->n` is greater than 1, the remaining `insn->n - 1` samples copied to userspace will be uninitialized kernel data. Another culprit is `vm80xx_ai_insn_read()` in the "vm80xx" driver. It never returns an error, even if it fails to fill the buffer. Fix it in `do_insn_ioctl()` and `do_insnlist_ioctl()` by making sure that uninitialized parts of the allocated buffer are zeroed before handling each instruction. Thanks to Arnaud Lecomte for their fix to `do_insn_ioctl()`. That fix replaced the call to `kmalloc_array()` with `kcalloc()`, but it is not always necessary to clear the whole buffer. | 19 Oct |
| CVE-2026-112147 | 5.5 medium | 0.00% | no CPE yet | A vulnerability was found in ai-genius-automations octoally up to 1.0.87. The impacted element is an unknown function of the file server/src/index.ts. Performing a manipulation of the argument path results in information disclosure. The attack may be initiated remotely. The exploit has been made public and could be used. Upgrading to version 1.0.88 is sufficient to resolve this issue. The patch is named 3898085c4e701be21377ebd9e91e10b941e47168. It is recommended to upgrade the affected component. | 19 Oct |
| CVE-2026-109403 | 5.5 medium | 0.00% | no CPE yet | A vulnerability was determined in conductor-oss conductor up to 3.32.5. Affected by this vulnerability is the function A2AService.getAgentCard of the file ai/src/main/java/org/conductoross/conductor/ai/a2a/A2AService.java of the component Agent Card Fetching. Executing a manipulation of the argument agentUrl can lead to server-side request forgery. The attack may be launched remotely. The exploit has been publicly disclosed and may be utilized. | 15 Oct |
| CVE-2026-45844 | 5.5 medium | 0.00% | linux, linux_kernel | In the Linux kernel, the following vulnerability has been resolved: netfilter: arp_tables: fix IEEE1394 ARP payload parsing Weiming Shi says: "arp_packet_match() unconditionally parses the ARP payload assuming two hardware addresses are present (source and target). However, IPv4-over-IEEE1394 ARP (RFC 2734) omits the target hardware address field, and arp_hdr_len() already accounts for this by returning a shorter length for ARPHRD_IEEE1394 devices. As a result, on IEEE1394 interfaces arp_packet_match() advances past a nonexistent target hardware address and reads the wrong bytes for both the target device address comparison and the target IP address. This causes arptables rules to match against garbage data, leading to incorrect filtering decisions: packets that should be accepted may be dropped and vice versa. The ARP stack in net/ipv4/arp.c (arp_create and arp_process) already handles this correctly by skipping the target hardware address for ARPHRD_IEEE1394. Apply the same pattern to arp_packet_match()." Mangle the original patch to always return 0 (no match) in case user matches on the target hardware address which is never present in IEEE1394. Note that this returns 0 (no match) for either normal and inverse match because matching in the target hardware address in ARPHRD_IEEE1394 has never been supported by arptables. This is intentional, matching on the target hardware address should never evaluate true for ARPHRD_IEEE1394. Moreover, adjust arpt_mangle to drop the packet too as AI suggests: In arpt_mangle, the logic assumes a standard ARP layout. Because IEEE1394 (FireWire) omits the target hardware address, the linear pointer arithmetic miscalculates the offset for the target IP address. This causes mangling operations to write to the wrong location, leading to packet corruption. To ensure safety, this patch drops packets (NF_DROP) when mangling is requested for these fields on IEEE1394 devices, as the current implementation cannot correctly map the FireWire ARP payload. This omits both mangling target hardware and IP address. Even if IP address mangling should be possible in IEEE1394, this would require to adjust arpt_mangle offset calculation, which has never been supported. Based on patch from Weiming Shi <[email protected]>. | 15 Oct |
| CVE-2026-8768 | 5.5 medium | 0.00% | vercel, ai | A vulnerability was found in vercel ai up to 3.0.97. The affected element is the function validateDownloadUrl of the file packages/provider-utils/src/download-blob.ts of the component provider-utils. The manipulation results in server-side request forgery. The attack can be launched remotely. The exploit has been made public and could be used. The vendor was contacted early about this disclosure but did not respond in any way. | 14 Oct |
| CVE-2026-101326 | 5.4 medium | 0.00% | kstover, ninja forms – contact form builder with calculators, quizzes, signatures & ai form builder | The The Ninja Forms – Contact Form Builder with Calculators, Quizzes, Signatures & AI Form Builder plugin for WordPress is vulnerable to arbitrary shortcode execution in all versions up to, and including, 3.15.5 This is due to the software allowing users to execute an action that does not properly validate a value before running do_shortcode. This makes it possible for authenticated attackers, with subscriber-level access and above, to execute arbitrary shortcodes. Exploitation requires that an administrator has configured at least one Ninja Forms field with its Default Value set to the {wp:user_display_name} merge tag on a publicly accessible page; a Subscriber need only set their display_name to a shortcode string via their profile and then visit that page. | Sat 31 Oct |
| CVE-2026-102923 | 5.4 medium | 0.00% | no CPE yet | SparkyFitness is a self-hosted, AI-powered health and fitness tracker for families. Prior to 1.6.0, SparkyFitnessServer/services/chatService.ts and SparkyFitnessServer/services/labelScanService.ts do not validate destinations supplied in custom AI service settings. An authenticated user saves a custom AI service URL and invokes an affected AI flow. The user can save a custom AI service URL and trigger an outbound request. Successful exploitation can send server-side POST requests with a controlled bearer value, probe internal services, and disclose the AI system prompt. This issue is fixed in version 1.6.0. | Wed 28 Oct |
| CVE-2026-40320 | 5.4 medium | 0.00% | giskard | Giskard is an open-source testing framework for AI models. In versions prior to 1.0.2b1, the ConformityCheck class rendered the rule parameter through Jinja2's default Template() constructor, silently interpreting template expressions at runtime. If check definitions are loaded from an untrusted source, a crafted rule string could achieve arbitrary code execution. Exploitation requires write access to a check definition and subsequent execution of the test suite. This issue has been fixed in giskard-checks version 1.0.2b1. | Mon 26 Oct |
| CVE-2026-54025 | 5.4 medium | 0.00% | librechat | LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.4-rc1, there is a vulnerability in LibreChat's markdown artifact preview pipeline. The marked library v15.0.12 does not HTML-escape double-quote characters in image alt text when a custom renderer falls through to the default renderer. LibreChat's generateMarkdownHtml function (in client/src/utils/markdown.ts) installs a custom image renderer that returns false for URLs passing the isSafeUrl allowlist check, which causes marked to fall back to its built-in renderer. That built-in renderer inserts the raw alt text into the alt="..." attribute without escaping double-quote characters. An attacker can craft an alt text such as " onload="payload to break out of the attribute and inject an arbitrary event handler. The resulting HTML is then assigned to document.getElementById('content').innerHTML inside the Sandpack preview iframe, causing the payload to execute in the victim's browser. This vulnerability is fixed in 0.8.4-rc1. | 25 Oct |
| CVE-2026-15085 | 5.4 medium | 0.00% | j-vee, ai_seo\/geo_analyzer | Improper Neutralization of Input During Web Page Generation ("Cross-site Scripting") vulnerability in Drupal AI SEO/GEO Analyzer allows Stored XSS. This issue affects AI SEO/GEO Analyzer versions: from 0.0.0 to 1.1.3. | 25 Oct |
| CVE-2026-55608 | 5.4 medium | 0.00% | n8n-mcp | n8n-MCP is an MCP server that provides AI assistants access to n8n node documentation, properties, and operations. Prior to 2.57.4, multi-tenant HTTP mode with ENABLE_MULTI_TENANT=true could allow an authenticated tenant to access default-scope workflow_versions backups instead of being confined to the tenant scope, exposing or deleting workflow-version backups from prior single-tenant deployments or migrations. This issue is fixed in version 2.57.4. | 24 Oct |
| CVE-2026-15069 | 5.4 medium | 0.00% | ibm, engineering_ai_hub | IBM Engineering AI Hub 1.0.0, 1.1.0, and 1.2.0 could allow a remote attacker to execute arbitrary script code due to improper neutralization of input during web page generation. | 24 Oct |
| CVE-2026-18266 | 5.4 medium | 0.00% | no CPE yet | Dify AI Workflow oauth_redirect_url Open Redirect Vulnerability. This vulnerability allows remote attackers to disclose sensitive information on affected installations of Dify. User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file. The specific flaw exists within the OAuth flow handling in the AppInitializer component. An attacker can force a redirection to a site that serves malicious content. An attacker can leverage this vulnerability to disclose information in the context of the application. Was ZDI-CAN-29196. | 24 Oct |
| CVE-2026-104001 | 5.4 medium | 0.00% | no CPE yet | The The Slim SEO – AI SEO Plugin, Lightweight, Fast & Automated plugin for WordPress is vulnerable to arbitrary shortcode execution in all versions up to, and including, 4.11.1 This is due to the software allowing users to execute an action that does not properly validate a value before running do_shortcode. This makes it possible for authenticated attackers, with subscriber-level access and above, to execute arbitrary shortcodes. A Subscriber-level user need only save the shortcode payload to their own Biographical Info field via the REST API or wp-admin profile page; the shortcode then executes on any unauthenticated front-end request that resolves that author's description, such as requests using the ?author= query parameter when auto_redirection is disabled. | 16 Oct |
| CVE-2026-21400 | 5.4 medium | 0.00% | intel, ai_reference_models | Protection mechanism failure for some Intel(R) AI Reference Models before version v3.4.1 within Ring 3: User Applications may allow an escalation of privilege. Unprivileged software adversary with a privileged user combined with a low complexity attack may enable escalation of privilege. This result may potentially occur via local access when attack requirements are present without special internal knowledge and requires passive user interaction. The potential vulnerability may impact the confidentiality (high), integrity (high) and availability (high) of the vulnerable system, resulting in subsequent system confidentiality (none), integrity (none) and availability (none) impacts. | 15 Oct |
Download these 749 as CSV · subscribe to this list.
EPSS scores: 2026-11-01. NVD cursor: 2026-10-29 00:22. The News page shows only notable CVEs (KEV, your stack, new criticals, EPSS over 50%, or active exploitation per SSVC); this table has all of them.