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-2026-90223 | 7.1 high | 0.00% | no CPE yet | In the Linux kernel, the following vulnerability has been resolved: nfc: llcp: bound SNL TLV parsing to the skb and add length checks nfc_llcp_recv_snl() walked the SNL TLV list using a u16 offset/length pair derived from skb->len, without bounding reads to the actual skb data. Three problems followed: - For a short frame (skb->len < LLCP_HEADER_SIZE), tlv_len underflowed. - The per-TLV header (type, length) was read without checking that two bytes remained. - A declared TLV length could run past the end of the buffer, and an SDREQ with length == 0 made "service_name_len = length - 1" underflow (size_t), driving an out-of-bounds read in the following strncmp() / nfc_llcp_sock_from_sn(). The SDRES case likewise read tlv[2]/tlv[3] without a length check. A nearby NFC device can reach this without authentication; LLCP link activation happens automatically after NFC-DEP. Walk the TLV list by pointer, bounded by skb_tail_pointer() over the linear skb data, and validate each TLV declared length before use. Add explicit length checks for SDREQ (>= 1) and SDRES (exactly 2). Found by 0sec automated security-research tooling (https://0sec.ai). | 19 Oct |
| CVE-2026-55787 | 7.1 high | 0.00% | no CPE yet | Flyto2 Core is an execution kernel for automation and AI-agent workflows. Prior to 2.26.3, validate_url_ssrf and is_private_ip in src/core/utils.py compare resolved addresses only against native private IPv4 and IPv6 ranges and do not unwrap IPv4-mapped, IPv4-compatible, 6to4, or NAT64 transition addresses. An authenticated workflow author can provide a transition-form literal URL through the Execution API POST /v1/execute route, http.get, or sibling guarded outbound-request modules, causing the guard to accept an address whose embedded IPv4 points to loopback, RFC 1918, or cloud metadata services. On a host that routes the selected transition form, the outbound request reaches the internal destination and returns the response body to the workflow caller, enabling read SSRF and disclosure of internal-service data or cloud credentials. This issue is fixed in version 2.26.3. | 16 Oct |
| CVE-2026-72415 | 7.1 high | 0.00% | no CPE yet | In the Linux kernel, the following vulnerability has been resolved: ASoC: SDCA: Validate written enum value in ge_put_enum_double() ge_put_enum_double() passes the user-supplied enumeration index item[0] to snd_soc_enum_item_to_val() without checking it against the number of items in the enum: ret = snd_soc_enum_item_to_val(e, item[0]); snd_soc_enum_item_to_val() indexes the heap-allocated e->values[] array with that index (e->values is set from a devm_kcalloc() of e->items entries), so a control write with an out-of-range item[0] reads past the end of the values buffer. The bounds check in snd_soc_dapm_put_enum_double() only runs afterwards, so it does not prevent the read here. Reject an out-of-range item before using it, matching the other enum put handlers. This issue was pointed out by the Sashiko AI review bot while reviewing a related enum-validation series: https://lore.kernel.org/all/[email protected]/ | 16 Oct |
| CVE-2026-72175 | 7.1 high | 0.00% | no CPE yet | In the Linux kernel, the following vulnerability has been resolved: fs/proc/task_mmu: fix make_uffd_wp_huge_pte() prot-update race Patch series "userfaultfd/pagemap: pre-existing fixes". These are pre-existing bug fixes that were carried at the front of the userfaultfd RWP working-set-tracking series up to v5 [1]. Per review feedback that fixes should not sit in the middle of a feature series, they are split out and sent on their own; the RWP series is reposted rebased on top of this. All six were flagged by the Sashiko AI review of the RWP series and carry independent of RWP, apply to mm-new directly, and carry Cc: stable@. 1: fs/proc/task_mmu: a missing huge_ptep_modify_prot_start() in make_uffd_wp_huge_pte() can lose hardware Dirty/Accessed updates when PAGEMAP_SCAN write-protects a hugetlb PTE. 2: fs/proc/task_mmu: pagemap_scan_hugetlb_entry() compares the range against HPAGE_SIZE rather than the hstate page size, so it never write-protects gigantic hugetlb pages. 3: fs/proc/task_mmu: PAGEMAP_SCAN with PM_SCAN_WP_MATCHING over an unpopulated hugetlb range self-deadlocks -- pagemap_scan_pte_hole() calls uffd_wp_range() while walk_hugetlb_range() holds the hugetlb vma lock for read, and hugetlb_change_protection() then takes it for write. Install the marker inline instead. 4: mm/huge_memory: change_non_present_huge_pmd() drops pmd_swp_uffd_wp on a device-private PMD permission downgrade, silently losing the uffd-wp marker. 5: userfaultfd: must_wait() applies pte_write() to a locklessly read PTE without checking pte_present(), so swap/migration entries decode random offset bits and a thread can stay parked on a stale fault. 6: userfaultfd: __VMA_UFFD_FLAGS feeds VMA_UFFD_MINOR_BIT (41) to mk_vma_flags() unconditionally, an out-of-bounds write into the single-word vma_flags_t on 32-bit. Build the mask from config-gated per-mode masks so an unavailable bit is never materialised. This patch (of 6): make_uffd_wp_huge_pte() arms the UFFD_WP bit on a present HugeTLB PTE by calling huge_ptep_modify_prot_commit() with a ptent snapshot that was fetched without the corresponding huge_ptep_modify_prot_start(). The start helper is what atomically clears the entry so the kernel-owned snapshot stays consistent until the commit; without it, the hardware may set Dirty or Accessed in the live PTE between the original read and the commit, and huge_ptep_modify_prot_commit() (whose generic implementation just calls set_huge_pte_at()) then writes the stale snapshot back over the live hardware bits, losing the update. The non-hugetlb sibling make_uffd_wp_pte() does this correctly via ptep_modify_prot_start() / ptep_modify_prot_commit(). Mirror that pattern for the present-PTE branch. The migration case stays as-is -- migration entries are non-present, so there's no hardware update to race against. | 16 Oct |
| CVE-2026-62676 | 7.1 high | 0.00% | no CPE yet | Omnigent is an open-source AI agent framework and meta-harness for orchestrating coding agents. Prior to 0.3.0, the shared shell-command parser in omnigent/policies/builtins/_shell.py fails to recognize combined interpreter flags, the timeout, nice, setsid, and stdbuf wrappers, command substitutions, and a single background control operator. A gated git push or gh write hidden with these forms produces no parsed operation, causing the github.py write_repos and write_branches allowlist and the working_dir.py workspace confinement policies to abstain and allow the command. An authenticated or prompt-injected agent can therefore push to an unauthorized repository or branch or escape the intended workspace. This issue is fixed in version 0.3.0. | 14 Oct |
| CVE-2026-110208 | 7.1 high | 0.00% | no CPE yet | Wukong AICRM through 20260610 contains a missing authorization vulnerability in the /systemConfig controller that allows any authenticated user to read the stored AI provider API key. Attackers can request GET /systemConfig/byType/ai for the cleartext key, or send a chosen apiUrl to /systemConfig/ai/test, forwarding the stored key to their host. | 14 Oct |
| CVE-2026-110209 | 7.1 high | 0.00% | no CPE yet | Stonewuu ai-fusion-video through 1.2.0 contains a missing authorization vulnerability that allows authenticated non-admin users to read stored secrets through unprotected ApiConfigController and SystemConfigController endpoints. Attackers can query the api-config get/list and system config endpoints to obtain AI provider API keys, proxy passwords and mail_password, enabling billing abuse and sending mail as the platform. | 13 Oct |
| CVE-2026-40803 | 7.1 high | 0.00% | no CPE yet | Unauthenticated Cross Site Scripting (XSS) in Jotform – AI Chatbot <= 3.8.2 versions. | 13 Oct |
| CVE-2026-94167 | 7.1 high | 0.00% | no CPE yet | Improper Neutralization of Input During Web Page Generation ('Cross-site Scripting') vulnerability in Extend Themes Kubio AI Page Builder kubio allows Reflected XSS.This issue affects Kubio AI Page Builder: from n/a through 2.9.3. | 9 Oct |
| CVE-2026-59704 | 7.1 high | 0.00% | no CPE yet | Cap's GET /api/video/ai endpoint fails to validate user ownership or membership before returning private video AI metadata including titles, summaries, and chapters. Authenticated attackers can supply arbitrary video IDs to read sensitive AI-generated content and trigger unauthorized AI generation that consumes the video owner's credits without consent. | 8 Oct |
| CVE-2026-105699 | 7.1 high | 0.00% | no CPE yet | Langflow is a tool for building and deploying AI-powered agents and workflows. From 1.6.8 until 1.9.1, Langflow authenticated access to the project identifier in a project-scoped MCP connection but did not authorize the resource URI supplied to resources/read. read_resource forwarded the attacker-controlled URI to handle_read_resource, which parsed a flow_id and filename and called storage_service.get_file without verifying that the flow belonged to the authenticated user or current project. A user with access to any project-scoped MCP endpoint could therefore request another user's flow-backed file, while global handle_list_resources and handle_list_tools behavior could disclose flow and file identifiers that made targeting easier. The vulnerability disclosed uploaded documents, structured data, prompts, and other private flow artifacts across tenants but did not modify victim files or stored flows. This issue is fixed in version 1.9.1. | 8 Oct |
| CVE-2026-54321 | 7.0 high | 0.00% | no CPE yet | Daytona is a secure and elastic infrastructure runtime for AI-generated code execution and agent workflows. From 0.101.0 until 0.184.0, sandbox previews that were switched from public to private could remain reachable without authentication for a short period after the change, due to a cached visibility state that was not invalidated when the sandbox's visibility changed. This vulnerability is fixed in 0.184.0. | 25 Oct |
| CVE-2026-44934 | 7.0 high | 0.00% | no CPE yet | A information disclosure when DEBUG loglevel is set in SUSE Rancher AI Agent 1.0 before 1.0.2 could leak API keys or LLM response text with potential sensitive data into logfiles, allowing local attackers to misuse respective gained data or credentials. | 25 Oct |
| CVE-2026-72718 | 7.0 high | 0.00% | no CPE yet | goose is general-purpose AI agent that runs on your machine. Prior to 1.44.0, the `goose review` command runs the system `git` executable to gather the diff for review without stripping attacker-controlled Git configuration. A malicious repository whose `.git/config` sets [`core] fsmonitor = <command>` causes Git to execute that command on the host during the index refresh performed by `git diff HEAD`. The command runs before goose contacts a model and without a submitted prompt, model call, tool approval, or trust prompt. The context-gathering Git process is not sandboxed and is outside goose's tool-permission model. Arbitrary commands run with the privileges and environment of the user running goose, allowing file access or modification and exfiltration of environment secrets and provider API keys. The vulnerable Git invocations are built by git_command() in crates/goose-cli/src/commands/review/handler.rs and are used by touched_files() and collect_diff() for `git diff --name-only HEAD` and `git diff HEAD`. This issue is fixed in version 1.44.0. | 24 Oct |
| CVE-2026-119711 | 6.9 medium | 0.00% | pipeshub-ai | PipesHub before 0.10.0 contains a rate limit bypass vulnerability in getClientIp() that allows unauthenticated attackers to evade per-IP throttling by spoofing X-Forwarded-For or X-Real-IP headers. Attackers can rotate header values to obtain fresh rate-limit buckets on authentication endpoints, enabling password spraying and OTP or password-reset email flooding. | Fri 30 Oct |
| CVE-2026-39422 | 6.9 medium | 0.00% | maxkb | MaxKB is an open-source AI assistant for enterprise. Versions 2.7.1 and below contain a Stored Cross-Site Scripting (XSS) vulnerability through the application name or icon fields when creating an application. When a victim visits the public chat interface (/ui/chat/{access_token}), the ChatHeadersMiddleware retrieves the application data and directly inserts the unescaped application name and icon into the HTML response via string replacement. This allows an attacker to execute arbitrary JavaScript in the victim's browser context. This issue has been fixed in version 2.8.0. | Mon 26 Oct |
| CVE-2026-39423 | 6.9 medium | 0.00% | maxkb | MaxKB is an open-source AI assistant for enterprise. Versions 2.7.1 and below contain an Eval Injection vulnerability in the Markdown rendering engine that allows any user capable of interacting with the AI chat interface to execute arbitrary JavaScript in the browsers of other users, including administrators, resulting in Stored Cross-Site Scripting (XSS). This issue has been fixed in version 2.8.0. | Mon 26 Oct |
| CVE-2026-12529 | 6.9 medium | 0.00% | no CPE yet | A security vulnerability has been detected in SourceCodester CET Automated Grading System with AI Predictive Analytics 1.0. Affected is an unknown function of the file /index.php of the component Student Self-Registration Endpoint. The manipulation leads to improper access controls. Remote exploitation of the attack is possible. | Mon 26 Oct |
| CVE-2026-45792 | 6.9 medium | 0.00% | rtk-ai, rtk | rtk filters and compresses command outputs before they reach your LLM context. Prior to 0.32.0, RTK (Rust Token Killer) improperly trusts project-local configuration files. RTK automatically loads .rtk/filters.toml from the working directory with highest priority and without user notification. An attacker can place a malicious filter file in a repository to apply regex-based modifications (e.g., strip_lines_matching) to shell command output before it is shown to the LLM, without any indication that the output has been modified. This allows attackers to selectively suppress or alter command output (including file contents, diffs, and security scan results) without detection, potentially concealing malicious code during AI-assisted development or review. This vulnerability is fixed in 0.32.0. | 25 Oct |
| CVE-2026-15746 | 6.9 medium | 0.00% | no CPE yet | Strands Agents is an open-source Python SDK for building and running AI agents. The strands-agents-tools package provides pre-built tools for use with the SDK, including the elasticsearch_memory tool for agent memory storage. We identified CVE-2026-15746, a server-side request forgery (SSRF) issue in the elasticsearch_memory tool. The tool exposed its connection parameters (es_url, cloud_id, api_key) as fields the large language model (LLM) could control through the tool schema. When a caller omitted the api_key parameter, the tool fell back to the operator's ELASTICSEARCH_API_KEY environment variable and sent it to whichever host the LLM specified. A crafted prompt could cause the tool to connect to a threat-actor-controlled server and disclose the operator's Elasticsearch API key in the Authorization header. We recommend you upgrade to strands-agents-tools version 0.7.0 or later. As a precautionary measure, we recommend all operators rotate their ELASTICSEARCH_API_KEY, even if there is no indication the credential was exposed. | 24 Oct |
| CVE-2026-110719 | 6.9 medium | 0.00% | no CPE yet | Portkey AI Gateway 1.14.0 through 1.15.2 contains a server-side request forgery vulnerability that allows unauthenticated attackers to reach loopback and Docker host services because isValidCustomHost trusts localhost-style hosts when TRUSTED_CUSTOM_HOSTS is unset. Attackers can send /v1 requests with a crafted x-portkey-custom-host header, such as via the ollama provider, to make the gateway call internal services and return their responses. | 20 Oct |
| CVE-2025-66405 | 6.9 medium | 0.00% | portkey, gateway | Portkey.ai Gateway is a blazing fast AI Gateway with integrated guardrails. Prior to 1.14.0, the gateway determined the destination baseURL by prioritizing the value in the x-portkey-custom-host request header. The proxy route then appends the client-specified path to perform an external fetch. This can be maliciously used by users for SSRF attacks. This vulnerability is fixed in 1.14.0. | 18 Oct |
| CVE-2026-105751 | 6.9 medium | 0.00% | docling | Docling simplifies document processing by parsing diverse formats and providing integrations with the generative AI ecosystem. From 2.107.0 until 2.120.3, docling/backend/opendocument_backend.py uses the xlink:href attribute value of a draw:image element as a filesystem path when the referenced part is not found in the document archive. The _image_ref_from_odf_image function reads that attacker-controlled path without a scheme check, extraction-directory confinement, or the enable_local_fetch setting used by other backends. Readable files that Pillow can decode as images are embedded in converted output, and other existing paths can be distinguished through the attempted read. This issue is fixed in 2.120.3. | 16 Oct |
| CVE-2026-108548 | 6.9 medium | 0.00% | no CPE yet | AstronRPA through 1.1.6 contains an authentication bypass vulnerability in the OpenResty gateway's auth_handler.lua that accepts any Bearer token without validation. Unauthenticated attackers can send arbitrary Bearer values to reach /api/resource/ and /api/rpa-ai-service/ routes and spoof X-User-Id or user_id headers to act as any user. | 16 Oct |
| CVE-2026-102671 | 6.9 medium | 0.00% | no CPE yet | The Joyland AI app accepts invalid SSL certificates in the invisible advertisement WebView by default. | 11 Oct |
| CVE-2026-102669 | 6.9 medium | 0.00% | no CPE yet | Joyland AI app does not verify hostnames, allowing a malicious host to connect or intercept chat messages. | 11 Oct |
| CVE-2025-62612 | 6.9 medium | 0.00% | fastgpt | FastGPT is an AI Agent building platform. Prior to version 4.11.1, in the workflow file reading node, the network link is not security-verified, posing a risk of SSRF attacks. This issue has been patched in version 4.11.1. | 8 Oct |
| CVE-2026-119343 | 6.8 medium | 0.00% | stirling-tools, stirling-pdf | Stirling-PDF is a locally hosted web application that facilitates various operations on PDF files. Prior to 3.0.0, POST /api/v1/ai/tools/create-pdf-from-html-agent accepts attacker-controlled htmlContent and passes it to WeasyPrint without CustomHtmlSanitizer, unlike the sanitized /api/v1/convert/html/pdf control path. HTTP and HTTPS resource references can make server-side requests to internal or cloud-metadata destinations, while file references can embed local content readable by the Stirling-PDF process into the generated PDF. The endpoint is reachable without credentials when security.enableLogin is false, and the flaw is distinct from earlier Stirling-PDF WeasyPrint vulnerabilities in older conversion paths. This issue is fixed in version 3.0.0. | Fri 30 Oct |
| CVE-2026-33990 | 6.8 medium | 0.00% | docker, model_runner | Docker Model Runner (DMR) is software used to manage, run, and deploy AI models using Docker. Prior to version 1.1.25, Docker Model Runner contains an SSRF vulnerability in its OCI registry token exchange flow. When pulling a model, Model Runner follows the realm URL from the registry's WWW-Authenticate header without validating the scheme, hostname, or IP range. A malicious OCI registry can set the realm to an internal URL (e.g., http://127.0.0.1:3000/), causing Model Runner running on the host to make arbitrary GET requests to internal services and reflect the full response body back to the caller. Additionally, the token exchange mechanism can relay data from internal services back to the attacker-controlled registry via the Authorization: Bearer header. This issue has been patched in version 1.1.25. For Docker Desktop users, enabling Enhanced Container Isolation (ECI) blocks container access to Model Runner, preventing exploitation. However, if the Docker Model Runner is exposed to localhost over TCP in specific configurations, the vulnerability is still exploitable. | Mon 26 Oct |
| CVE-2026-49988 | 6.8 medium | 0.00% | yamadashy, repomix | Repomix is a tool that packs repositories into AI-friendly files. Prior to 1.14.1, the Repomix MCP server attach_packed_output and read_repomix_output flow can register and read arbitrary local .json, .txt, .md, or .xml files without the file_system_read_file runSecretLint() safety check or Repomix packed-output validation, allowing MCP callers to bypass the local file-read secret-scanning boundary. This issue is fixed in version 1.14.1. | 24 Oct |
| CVE-2026-112014 | 6.8 medium | 0.00% | no CPE yet | LobsterAI before 2026.10.16 contains a hardcoded password vulnerability that allows attackers to decrypt exported provider API keys encrypted with the constant 'lobsterai-APP'. Attackers who obtain a providers export file can derive the AES-GCM key via PBKDF2 using the embedded salt and recover every third-party AI provider API key. | 20 Oct |
| CVE-2025-12351 | 6.8 medium | 0.00% | no CPE yet | Honeywell S35 Series Cameras contains an authorization bypass Vulnerability through User controller key. An attacker could potentially exploit this vulnerability, leading to Privilege Escalation to admin privileged functionalities . Honeywell also recommends updating to the most recent version of this product, service or offering (S35 Pinhole/Kit Camera to version 2025.08.28, S35 AI Fisheye & Dual Sensor/Micro Dome/Full Color Eyeball & Bullet Camera to version 2025.08.22, S35 Thermal Camera to version 2025.08.26). | 18 Oct |
| CVE-2025-8084 | 6.8 medium | 0.00% | no CPE yet | The AI Engine plugin for WordPress is vulnerable to Server-Side Request Forgery in all versions up to, and including, 3.1.8 via the rest_helpers_create_images function. This makes it possible for authenticated attackers, with Editor-level access and above, to make web requests to arbitrary locations originating from the web application and can be used to query and modify information from internal services. On Cloud instances, this issue allows for metadata retrieving. | 18 Oct |
| CVE-2026-107289 | 6.8 medium | 0.00% | pydantic, pydantic_ai | Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.56.0 until 1.107.6 and 2.44.0, applications that opt attacker-influenced URLs into local network access through FileUrl with force_download='allow-local' or web_fetch_tool with allow_local_urls=True can bypass the cloud-metadata blocklist by appending an IPv6 zone identifier to an IPv6 metadata address. IPv6Address equality and hashing include the zone identifier, so the blocklist comparison fails even though the network stack ignores the zone on a non-link-local destination and reaches the metadata service, potentially exposing cloud IAM credentials. The opt-in settings are disabled by default, and the issue requires an IPv6-enabled environment. This issue is fixed in versions 1.107.6 and 2.44.0. | 13 Oct |
| CVE-2026-22551 | 6.7 medium | 0.00% | eclipse, theia | In Eclipse Theia versions prior to 1.71.0, the AI chat rendered Markdown image tags from AI responses, triggering HTTP requests to arbitrary external URLs without restriction. Combined with prompt injection in a malicious workspace, an attacker could induce the AI agent to construct image URLs encoding sensitive information from the workspace or conversation context, exfiltrating it to attacker-controlled servers. The workspace trust enforcement introduced in v1.71.0 mitigates the documented attack chain by disabling AI features in untrusted workspaces. | Mon 26 Oct |
| CVE-2026-44725 | 6.6 medium | 0.00% | no CPE yet | EMQX is a scalable and reliable MQTT broker for AI, IoT, IIoT, and connected vehicles. Prior to versions 5.8.11, 5.9.3, 5.10.4, 6.0.3, 6.1.2, and 6.2.1, the plugin-install REST API and dashboard upload accepted stale grants created with emqx ctl plugins allow because there was no five-minute grant lifetime or SHA-256 package binding. An attacker with a compromised dashboard administrator credential or API key with plugin-install permission who finds a stale allowed name and version can upload attacker-controlled bytes under the allowed .tar.gz filename through POST /api/v5/plugins/install or the dashboard plugin upload. The broker then installs and runs attacker-controlled Erlang code with the privileges of the EMQX process. This issue is fixed in versions 5.8.11, 5.9.3, 5.10.4, 6.0.3, 6.1.2, and 6.2.1. | Mon 26 Oct |
| CVE-2026-44192 | 6.6 medium | 0.00% | no CPE yet | A flaw was found in the Ansible Lightspeed Model Context Protocol (MCP) server. This vulnerability, known as path traversal, allows an attacker to manipulate an AI agent through indirect prompt injection. By doing so, the attacker can cause the server to write files to unauthorized locations on the user's system. This can result in the exposure of sensitive host information and enable the attacker to execute malicious commands, potentially leading to a full system compromise. | 24 Oct |
| CVE-2026-116691 | 6.5 medium | 0.00% | no CPE yet | MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. From 2.19.0 until 3.17.0, mlflow.crewai.autolog() records CrewAI Agent objects without consistently sanitizing nested credential-bearing values. mlflow/crewai/autolog.py::_get_agent_attributes() converts the Agent llm field to a string that includes the provider API key, while mlflow/crewai/autolog.py::_construct_full_inputs() expands the Agent object into mlflow.spanInputs. Applications can therefore send plaintext provider credentials to the MLflow trace backend or a configured OTLP destination, where trace readers can obtain them. This issue is fixed in version 3.17.0. | Wed 28 Oct |
| CVE-2026-40150 | 6.5 medium | 0.00% | praison, praisonaiagents | PraisonAIAgents is a multi-agent teams system. Prior to 1.5.128, the web_crawl() function in praisonaiagents/tools/web_crawl_tools.py accepts arbitrary URLs from AI agents with zero validation. No scheme allowlisting, hostname/IP blocklisting, or private network checks are applied before fetching. This allows an attacker (or prompt injection in crawled content) to force the agent to fetch cloud metadata endpoints, internal services, or local files via file:// URLs. This vulnerability is fixed in 1.5.128. | Mon 26 Oct |
| CVE-2026-40980 | 6.5 medium | 0.00% | vmware, spring_ai | In Spring AI, a malicious PDF file can be crafted that triggers the allocation of unreasonable amounts of memory when handled by `ForkPDFLayoutTextStripper`. 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) | Mon 26 Oct |
| CVE-2026-42867 | 6.5 medium | 0.00% | langflow | Langflow is a tool for building and deploying AI-powered agents and workflows. Prior to 1.9.0, Langflow is vulnerable to Path Traversal in the Knowledge Bases API (POST /api/v1/knowledge_bases). This occurs because user-supplied knowledge base names are used directly to create file paths without proper sanitization or containment checks. An authenticated attacker can exploit this flaw to create directories and write files anywhere on the server's filesystem. This vulnerability is fixed in 1.9.0. | Mon 26 Oct |
| CVE-2026-54324 | 6.5 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, a cross-tenant authorization flaw in Daytona's notification WebSocket gateway allowed any authenticated user to subscribe to another organization's realtime notification channel and passively receive that organization's events. This vulnerability is fixed in 0.185.0. | 25 Oct |
| CVE-2026-54024 | 6.5 medium | 0.00% | librechat | LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.4-rc1, the fix for CVE-2024-11171 (commit bb58a2d0) added limits: { fileSize } to createMulterInstance() in the file upload routes. However, the POST /api/convos/import endpoint uses a separate multer instance that was never updated with the same limits configuration. Combined with the application-level size check being disabled by default (the CONVERSATION_IMPORT_MAX_FILE_SIZE_BYTES env var is commented out in .env.example), an authenticated user can upload arbitrarily large files to exhaust server disk space and memory. This vulnerability is fixed in 0.8.4-rc1. | 25 Oct |
| CVE-2026-54027 | 6.5 medium | 0.00% | librechat | LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.4-rc1, the POST /api/files/images endpoint allows any authenticated user to upload files into any agent's tool_resources (e.g., context, execute_code) without verifying ownership or EDIT permission on the target agent. A permission check was added to the POST /api/files route in a previous patch, but the image upload route was never updated with the same check. An attacker can simply use the image endpoint instead of the file endpoint to bypass the authorization entirely. This vulnerability is fixed in 0.8.4-rc1. | 25 Oct |
| CVE-2026-54029 | 6.5 medium | 0.00% | librechat | LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.4-rc1, the DELETE /api/messages/:conversationId/:messageId endpoint allows any authenticated user to delete any other user's messages. The validateMessageReq middleware only validates that the conversationId belongs to the requesting user, but the handler calls deleteMessages({ messageId }) using only the messageId as the MongoDB filter — without adding a user constraint. An attacker provides their own valid conversationId (to pass validation) and the victim's messageId (to target deletion), resulting in permanent, irrecoverable message deletion. This vulnerability is fixed in 0.8.4-rc1. | 25 Oct |
| CVE-2026-54033 | 6.5 medium | 0.00% | librechat | LibreChat is an enhanced ChatGPT clone that supports multiple AI providers. Prior to 0.8.4-rc1, LibreChat allows users to configure custom OpenAI-compatible API endpoints by setting a baseURL. This URL is used to construct HTTP requests without any SSRF validation — no private IP check, no scheme restriction, no DNS pinning. An authenticated user can set baseURL to internal network addresses. This vulnerability is fixed in 0.8.4-rc1. | 25 Oct |
| CVE-2026-13437 | 6.5 medium | 0.00% | devolutions, powershell_universal | Insertion of sensitive information into sent data in the AI Agent job API in Devolutions PowerShell Universal 2026.2.0 allows an authenticated user with AI Agent read access to obtain reusable, potentially higher-privileged authentication tokens via App Tokens serialized in plaintext in job API responses. | 25 Oct |
| CVE-2026-15154 | 6.5 medium | 0.00% | redhat, openshift_ai | A flaw was found in `guardrails-detectors`, a component of Red Hat OpenShift AI. This vulnerability, known as Regular Expression Denial of Service (ReDoS), allows a remote attacker to provide specially crafted regular expressions to the public detection API. This can cause catastrophic backtracking, leading to a worker process consuming 100% CPU indefinitely and resulting in a denial of service for the entire guardrails-mediated LLM pipeline. | 25 Oct |
| CVE-2026-58481 | 6.5 medium | 0.00% | network-ai | Network-AI is a TypeScript/Node.js multi-agent orchestrator. Prior to version 5.12.2, `AgentRuntime` promises scoped file access under a configured sandbox `basePath`, but its path containment checks use raw string prefix tests. A sandbox base such as `/tmp/network-ai-sandbox` also matches a sibling path such as `/tmp/network-ai-sandbox_evil/secret.txt`. An agent/user that can call `AgentRuntime.readFile()` or `AgentRuntime.listDir()` can read or list files outside the intended sandbox when the target path is in a sibling directory sharing the base path prefix. This breaks the documented sandbox boundary. The issue is fixed in v5.12.2. `SandboxPolicy.resolvePath()` and `isPathAllowed()` now use separator-anchored prefix checks (`resolved === base || resolved.startsWith(base + path.sep)`) for both the allow-list and block-list. A sibling directory that merely shares a name prefix (e.g. `/srv/app-evil` vs base `/srv/app`) is no longer treated as in-scope. | 24 Oct |
| CVE-2026-87936 | 6.5 medium | 0.00% | no CPE yet | Improper Neutralization of Special Elements used in an SQL Command ('SQL Injection') vulnerability in Drupal amazee.ai Private AI Provider allows SQL Injection. This issue affects amazee.ai Private AI Provider versions: from 0.0.0 to 1.3.7, from 1.4.0 to 1.4.3. | 22 Oct |
Download these 743 as CSV · subscribe to this list.
EPSS scores: 2026-10-31. 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.