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-2026-13437 | 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 |
| CVE-2026-17557 | 6.5 medium | 0.00% | no CPE yet | A flaw was found in the `lm-evaluation-harness` component of Red Hat OpenShift AI. An authenticated user, able to submit an evaluation job, could exploit this by providing a malicious model URL in the `JobSpec`. This vulnerability allows for the disclosure of the tenant's model authentication secret to an attacker-controlled external listener. Additionally, it enables Server-Side Request Forgery (SSRF) attacks against internal network endpoints reachable from the evaluation job pod. | 22 Oct |
| CVE-2026-86332 | 6.5 medium | 0.00% | no CPE yet | A flaw was found in odh-dashboard in Red Hat OpenShift AI. The backend-for-frontend route GET /api/nim-serving/:nimResource reads Kubernetes Secrets using the dashboard service account and returns the full Secret object, including .data, without an authorization check. Any authenticated dashboard user can retrieve the cluster NVIDIA NGC API key Secret (apiKeySecret) and the NIM image pull secret (nimPullSecret). Create and delete of the same NIM credential are admin-gated; the read path is not. This is missing authorization (CWE-862) and insufficiently protected credentials (CWE-522). It is distinct from CVE-2026-5483 (service-account token leak in the Kubernetes client response wrapper on the same route) and CVE-2026-16456 (odh-model-controller cross-namespace confused deputy). | 22 Oct |
| CVE-2025-67743 | 6.5 medium | 0.00% | learningcircuit, local_deep_research | Local Deep Research is an AI-powered research assistant for deep, iterative research. In versions from 1.3.0 to before 1.3.9, the download service (download_service.py) makes HTTP requests using raw requests.get() without utilizing the application's SSRF protection (safe_requests.py). This can allow attackers to access internal services and attempt to reach cloud provider metadata endpoints (AWS/GCP/Azure), as well as perform internal network reconnaissance, by submitting malicious URLs through the API, depending on the deployment and surrounding controls. This issue has been patched in version 1.3.9. | 19 Oct |
| CVE-2025-13378 | 6.5 medium | 0.00% | no CPE yet | The AI ChatBot with ChatGPT and Content Generator by AYS plugin for WordPress is vulnerable to Server-Side Request Forgery in all versions up to, and including, 2.7.0 via the ays_chatgpt_pinecone_upsert function. This makes it possible for unauthenticated attackers to make web requests to arbitrary locations originating from the web application and can be used to query and modify information from internal services. | 19 Oct |
| CVE-2025-12174 | 6.5 medium | 0.00% | no CPE yet | The Directorist: AI-Powered Business Directory Plugin with Classified Ads Listings plugin for WordPress is vulnerable to unauthorized access due to a missing capability check on the 'directorist_prepare_listings_export_file' and 'directorist_type_slug_change' AJAX actions in all versions up to, and including, 8.5.2. This makes it possible for authenticated attackers, with Subscriber-level access and above, to export listing details and change the directorist slug. | 18 Oct |
| CVE-2025-13359 | 6.5 medium | 0.00% | taxopress | The Tag, Category, and Taxonomy Manager – AI Autotagger with OpenAI plugin for WordPress is vulnerable to time-based SQL Injection via the "getTermsForAjax" function in all versions up to, and including, 3.40.1. This is due to insufficient escaping on the user supplied parameters and lack of sufficient preparation on the existing SQL query. This makes it possible for authenticated attackers, with contributor level access and above, to append additional SQL queries into already existing queries that can be used to extract sensitive information from the database granted they have metabox access for the taxonomy (enabled by default for contributors). | 18 Oct |
| CVE-2026-109382 | 6.5 medium | 0.00% | no CPE yet | The WSP MCP AI Agents Connector plugin for WordPress is vulnerable to Sensitive Information Disclosure in versions up to, and including, 2.9.5. This is due to the wsp_get_posts MCP tool being registered with an empty capability requirement while the MCP endpoint's authenticate() method accepts any logged-in user, and the wsp_execute_get_posts() callback honouring a caller-supplied status=all argument that expands into a WP_Query over 'publish', 'draft', 'pending' and 'future' posts with no author or ownership filter. This makes it possible for authenticated attackers, with Subscriber-level access and above, to read the titles, URLs, authors, categories, tags and 40-word excerpts of every unpublished (draft/pending/scheduled) post on the site. | 16 Oct |
| CVE-2025-13922 | 6.5 medium | 0.00% | no CPE yet | The Tag, Category, and Taxonomy Manager – AI Autotagger with OpenAI plugin for WordPress is vulnerable to time-based blind SQL Injection via the 'existing_terms_orderby' parameter in the AI preview AJAX endpoint in all versions up to, and including, 3.40.1. This is due to insufficient escaping on user-supplied parameters and lack of SQL query parameterization. This makes it possible for authenticated attackers, with Contributor-level access and above who have AI metabox permissions, to append additional SQL queries into already existing queries that can be used to extract sensitive information from the database, cause performance degradation, or enable data inference through time-based techniques. | 15 Oct |
| CVE-2026-5163 | 6.5 medium | 0.00% | mattermost, mattermost_server | Mattermost versions 11.5.x <= 11.5.1 fail to verify channel membership when processing AI-assisted message rewrites which allows an authenticated attacker to read the content of threads in private channels and direct messages they do not have access to via a crafted request to the post rewrite endpoint.. Mattermost Advisory ID: MMSA-2026-00645 | 14 Oct |
| CVE-2026-107287 | 6.5 medium | 0.00% | pydantic, pydantic_ai | Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.77.0 until 1.107.7 and 2.52.0, the local web_fetch_tool and the WebFetch local fallback can consume excessive CPU and memory during HTML-to-Markdown conversion of attacker-controlled HTML containing deeply nested block elements. Conversion repeatedly reprocesses accumulated text and can greatly expand intermediate output before the returned-content limit is applied, allowing a model-directed fetch to delay other work in the process. Provider-native web fetching is not affected. This issue is fixed in versions 1.107.7 and 2.52.0. | 14 Oct |
| CVE-2026-107294 | 6.5 medium | 0.00% | pydantic, pydantic_ai | Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.77.0 until 1.107.2 and 2.24.0, web_fetch_tool, the WebFetch local fallback, and remote FileUrl media downloads buffer the complete HTTP response body before enforcing content-size controls. An attacker-influenced URL can stream an arbitrarily large response that exhausts process memory and crashes the worker; affected media types include ImageUrl, DocumentUrl, VideoUrl, and AudioUrl. SSRF protections remain effective, and the impact is limited to availability. This issue is fixed in versions 1.107.2 and 2.24.0. | 13 Oct |
| CVE-2026-104915 | 6.5 medium | 0.00% | no CPE yet | The Academy LMS – AI Course Builder, Quizzes, Certificates & eLearning plugin for WordPress is vulnerable to authorization bypass in all versions up to, and including, 4.0.3. This is due to the plugin not properly verifying that a user is authorized to perform an action. This makes it possible for authenticated attackers, with custom-level access and above, to permanently delete arbitrary Academy lesson comments and their replies authored by any user, including administrators, across courses they do not instruct. This is exploitable by any user registered as an Academy instructor for at least one course, as they can supply their own course_id to pass the instructor check while targeting comments belonging to entirely different courses. | 13 Oct |
| CVE-2026-105749 | 6.5 medium | 0.00% | docling | Docling simplifies document processing by parsing diverse formats and providing integrations with the generative AI ecosystem. From 2.0.0 until 2.131.0, the HTML, JATS, OpenDocument spreadsheet, and BoxNote backends, including docling/backend/html_backend.py, docling/backend/jats_backend.py, and docling/backend/boxnote_backend.py, accept the rowspan and colspan attribute values without an upper bound and execute loops or allocate a table grid proportional to the declared span. A very small document can therefore cause sustained CPU use or multi-gigabyte memory allocation, and the document_timeout setting does not interrupt the single backend conversion call. Export through the TableData.grid property can further materialize the oversized grid. This issue is fixed in 2.131.0. | 8 Oct |
| CVE-2026-116546 | 6.4 medium | 0.00% | wpjobportal, wp job portal – ai-powered recruitment system for company or job board website | The WP Job Portal – AI-Powered Recruitment System for Company or Job Board website plugin for WordPress is vulnerable to Stored Cross-Site Scripting via the 'degreetitle' parameter in all versions up to, and including, 2.7.0 due to insufficient input sanitization and output escaping. This makes it possible for authenticated attackers, with subscriber-level access and above, to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page. The plugin's self-registration flow allows any visitor to create a Subscriber-backed Employer account, meaning this vulnerability is effectively exploitable by any user who can register on the site. | 13 h ago |
| CVE-2026-116535 | 6.4 medium | 0.00% | rainbowgeek, seopress – ai seo plugin & on-site seo | The SEOPress – AI SEO Plugin & On-site SEO plugin for WordPress is vulnerable to Stored Cross-Site Scripting via the 'display_name' parameter in all versions up to, and including, 10.3 due to insufficient input sanitization and output escaping. This makes it possible for authenticated attackers, with subscriber-level access and above, to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page. This is exploitable by any Subscriber who sets their display_name to a string containing the literal %%archive_title%% token alongside HTML event-handler payloads, as WordPress stores display names without HTML filtering and the plugin substitutes the token after escaping has already occurred. | 13 h ago |
| CVE-2026-108366 | 6.4 medium | 0.00% | no CPE yet | The easy.jobs – AI powered Job Listing, Job Board, Career Page, Recruitment & Hiring Solution plugin for WordPress is vulnerable to Authenticated (Contributor+) Stored Cross-Site Scripting via the 'icon' attribute of the easyjobs/job-list Gutenberg block in all versions up to, and including, 2.8.2 due to insufficient input sanitization and output escaping on user supplied attributes. This makes it possible for authenticated attackers, with contributor-level access and above, to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page. | Wed 28 Oct |
| CVE-2026-109385 | 6.4 medium | 0.00% | no CPE yet | The UiChemy – Convert Figma, Prompts and AI Builds to Elementor, Gutenberg and Bricks plugin for WordPress is vulnerable to Stored Cross-Site Scripting via the 'actions[save_builder][data][elements][...][settings][slot_{N}_link][url]' parameter in all versions up to, and including, 5.3.0 due to insufficient input sanitization and output escaping. This makes it possible for authenticated attackers, with contributor-level access and above, to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page. The injected javascript: payload fires on click rather than on page load, making exploitation contingent on a privileged user such as an Editor or Administrator clicking the attacker-crafted link while reviewing or previewing the Contributor's post. | Wed 28 Oct |
| CVE-2026-40346 | 6.4 medium | 0.00% | nocobase | NocoBase is an AI-powered no-code/low-code platform for building business applications and enterprise solutions. Prior to version 2.0.37, NocoBase's workflow HTTP request plugin and custom request action plugin make server-side HTTP requests to user-provided URLs without any SSRF protection. An authenticated user can access internal network services, cloud metadata endpoints, and localhost. Version 2.0.37 contains a patch. | Mon 26 Oct |
| CVE-2026-55370 | 6.4 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's existing TOTP verification accepted a successfully used TOTP code again while the code remained inside the RFC 6238 acceptance window because the verifier used otplib's stateless check with window = 1 and did not persist or compare the accepted TOTP time-step counter. An attacker who has the victim's first factor and captures a live TOTP value can replay that value to satisfy MFA during the same acceptance window. This issue is fixed in version 1.41.0. | 25 Oct |
| CVE-2026-57413 | 6.4 medium | 0.00% | no CPE yet | Server-Side Request Forgery (SSRF) vulnerability in bdthemes Instant Image Generator ai-image allows Server Side Request Forgery.This issue affects Instant Image Generator: from n/a through <= 2.1.4. | 24 Oct |
| CVE-2026-56678 | 6.4 medium | 0.00% | no CPE yet | 9Router is an AI router & token saver. Prior to 0.5.6, the Kiro API-key validation endpoint POST /api/oauth/kiro/api-key builds an upstream URL using a user-controlled region value, allowing an authenticated attacker to supply a crafted region such as kiro-canary.local:8443# and cause 9Router to send the Kiro validation request to an attacker-controlled host while forwarding the submitted Kiro API key as an Authorization header. This issue is fixed in version 0.5.6. | 24 Oct |
| CVE-2026-47361 | 6.4 medium | 0.00% | no CPE yet | In versions of the Datadog Android application prior to v541-5.9.2, BubbleChatActivity is exported with no permission guard and accepts a SEND intent with a caller-supplied conversation_id. When the activity closes and no in-process session matches that ID, it unconditionally cancels notification ID 9201 (the Bits AI chat notification), with no check on the caller's identity or ownership of the conversation. This requires a malicious application co-installed on the victim's device. Impact: A co-installed application can silently dismiss the victim's Bits AI chat notification. No chat content is exposed; conversation data remains server-authentication gated and is never returned to the caller. | 23 Oct |
| CVE-2026-75130 | 6.4 medium | 0.00% | no CPE yet | Context7 through 2.1.2 contains a prompt injection vulnerability that allows attackers to execute malicious instructions in connected AI coding agents by injecting unsanitized content through the Custom AI Instructions feature served via the MCP server. Attackers can poison the custom instructions to exfiltrate credentials from environment files to an attacker-controlled service and perform destructive file deletion on the victim's machine when the agent makes a routine library documentation request. | 22 Oct |
| CVE-2026-109985 | 6.4 medium | 0.00% | no CPE yet | The xSpeed Cache: AI-Powered Performance Hub with MCP, Caching & CDN plugin for WordPress is vulnerable to Stored Cross-Site Scripting via the 'content' parameter in all versions up to, and including, 1.4.2 due to insufficient input sanitization and output escaping. This makes it possible for authenticated attackers, with contributor-level access and above, to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page. | 19 Oct |
| CVE-2026-103480 | 6.4 medium | 0.00% | no CPE yet | The Tourfic – AI Powered Travel Booking, Hotel Booking & Car Rental plugin for WordPress is vulnerable to Stored Cross-Site Scripting via the 'type' parameter in all versions up to, and including, 3.0.1 due to insufficient input sanitization and output escaping. This makes it possible for authenticated attackers, with contributor-level access and above, to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page. The wp_kses_post filter applied on save does not strip the payload because angle-bracket characters are not required, allowing the unescaped attribute value to survive storage and fire as a live event handler in the unauthenticated (guest) rendering branch. | 16 Oct |
| CVE-2026-97470 | 6.4 medium | 0.00% | no CPE yet | The Tourfic – AI Powered Travel Booking, Hotel Booking & Car Rental plugin for WordPress is vulnerable to Stored Cross-Site Scripting via Comment Content via unfiltered_html Grant to Customer Role in all versions up to, and including, 3.0.1 due to insufficient input sanitization and output escaping. This makes it possible for authenticated attackers, with subscriber-level access and above, to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page. This requires WooCommerce to be active, but since the customer role is freely self-registerable without admin approval, exploitation is effectively possible by any unauthenticated visitor who registers a WooCommerce account. While the plugin's own listing-page templates sanitize output with wp_kses_post(), the stored payload is returned unsanitized via the unauthenticated GET /wp-json/wp/v2/comments REST endpoint (content.rendered) and post/comment RSS feeds, exposing any headless client, feed reader, or third-party widget that renders comment content. | 16 Oct |
| CVE-2026-93535 | 6.4 medium | 0.00% | no CPE yet | The Dark Mode – Improve Accessibility with AI Powered Dark Theme plugin for WordPress is vulnerable to Stored Cross-Site Scripting via the 'icon_light' Shortcode Attribute in all versions up to, and including, 5.3.17 due to insufficient input sanitization and output escaping. This makes it possible for authenticated attackers, with Contributor-level access and above, to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page. | 16 Oct |
| CVE-2026-100185 | 6.4 medium | 0.00% | no CPE yet | The Directorist: AI-Powered Business Directory, Listings & Classified Ads plugin for WordPress is vulnerable to Stored Cross-Site Scripting via 'phone' Listing Field in all versions up to, and including, 8.10.2.1 due to insufficient input sanitization and output escaping. This makes it possible for authenticated attackers, with subscriber-level access and above, to inject arbitrary web scripts in pages that will execute whenever a user accesses an injected page. This is possible because the plugin grants the edit_at_biz_dirs and publish_at_biz_dirs capabilities to the subscriber role, allowing any subscriber-level user to submit and publish listings without moderation. | 15 Oct |
| CVE-2026-107292 | 6.4 medium | 0.00% | pydantic, pydantic_ai | Pydantic AI is a Python agent framework for building applications and workflows with Generative AI. From 1.34.0 until 2.30.0, the Agent.to_web() and clai web development chat server does not validate the Host header, allowing a website visited by a developer to use DNS rebinding to reach a loopback-hosted agent as a same-origin service. The hostile page can read the served UI and submit chat requests that execute agent tools with the local process's privileges and credentials, causing data disclosure or unwanted side effects. Binding to localhost, Origin checks, and CSRF tokens do not prevent the same-origin DNS rebinding path. This issue is fixed in versions 1.107.5 and 2.30.0. | 14 Oct |
| CVE-2025-11361 | 6.4 medium | 0.00% | no CPE yet | The Gutenberg Essential Blocks – Page Builder for Gutenberg Blocks & Patterns plugin for WordPress is vulnerable to Server-Side Request Forgery in all versions up to, and including, 5.7.1 via the eb_save_ai_generated_image function. This makes it possible for authenticated attackers, with Author-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. | 9 Oct |
| CVE-2026-120204 | 6.3 medium | 0.00% | deepseek-ai, deepinfer | DeepInfer through 0.1.0 contains a path traversal vulnerability in the unauthenticated /profile endpoint that allows remote attackers to create arbitrary directories via the output_dir parameter. Attackers, including cross-site GET requests from an operator's browser, can write chrome trace files to exhaust disk or inodes and use error messages as a filesystem oracle. | Sat 31 Oct |
| CVE-2025-13327 | 6.3 medium | 0.00% | astral-sh, red hat, uv, red hat openshift ai 3.0, red hat ai inference server | A flaw was found in uv. This vulnerability allows an attacker to execute malicious code during package resolution or installation via specially crafted ZIP (Zipped Information Package) archives that exploit parsing differentials, requiring user interaction to install an attacker-controlled package. | Thu 29 Oct |
| CVE-2026-116575 | 6.3 medium | 0.00% | no CPE yet | Mem0 provides persistent memory infrastructure for AI agents and applications. Prior to Python SDK 2.0.18 and TypeScript SDK 3.1.6, _build_session_scope() in mem0/memory/main.py and Memory.buildSessionScope() in the TypeScript SDK join caller-controlled user_id, agent_id, and run_id values using unescaped & and = delimiters. A caller can embed another scope component in a controlled identifier so that distinct tenant tuples produce the same session_scope key. Memory.add(), AsyncMemory.add(), and the REST POST /memories flow use that key to read and save recent raw messages, allowing cross-tenant conversation disclosure and eviction or interleaving of another tenant's history. This issue is fixed in Python SDK 2.0.18 and TypeScript SDK 3.1.6. | Tue 27 Oct |
| CVE-2026-15044 | 6.3 medium | 0.00% | no CPE yet | A flaw was found in the TrustyAI Service Operator. When deploying services like gorch or NemoGuardrails, if a specific security setting is not enabled, these services can expose their communication channels without requiring users to prove their identity. This allows any other program within the cluster to access the AI guardrails and orchestrator without proper authorization. An attacker could exploit this to gain unauthorized access to sensitive information and potentially make limited changes to the AI models. | Mon 26 Oct |
| CVE-2026-54601 | 6.3 medium | 0.00% | no CPE yet | FastGPT is an open source AI knowledge base platform. From 4.14.17 to before 4.15.0-beta4, FastGPT allows an authenticated tenant user to call POST /api/core/dataset/collection/create/reTrainingCollection in a way that persists a server-owned datasetId value from another tenant. This creates mixed dataset objects and downstream dataset, collection, and training endpoints then make authorization decisions from inconsistent ownership anchors, allowing cross-tenant read, update, and delete access when mixed object ids are known. This issue is fixed in version 4.15.0-beta4. | 25 Oct |
| CVE-2026-57476 | 6.3 medium | 0.00% | deloitte, ai_assist_for_customer | Deloitte AI Assist for Customer exposed unauthenticated API endpoints that allowed an attacker with knowledge of additional parameters to read from or inject content into the retrieval-augmented generation (RAG) corpus. On 2026-03-25, AI Assist for Customer restricted network access and enforced authentication for the previously exposed endpoints. | 25 Oct |
| CVE-2026-61646 | 6.3 medium | 0.00% | no CPE yet | FastGPT is a knowledge-based AI application platform. Prior to 4.15.0-beta5, FastGPT's shared SSRF guard validates only the initial request URL before handing the request to axios, and axios follows redirects by default. An authenticated workflow user can configure an HTTP request node to call an attacker-controlled public URL that redirects to cloud metadata, loopback, or internal services that the guard would block on direct request, and the HTTP node returns the response body to the workflow caller. This issue is fixed in version 4.15.0-beta5. | 24 Oct |
| CVE-2026-64650 | 6.3 medium | 0.00% | no CPE yet | The `@ai-sdk/harness-opencode` tool is an HarnessV1 adapter backed by @openai/codex-sdk, which drives the codex command line interface. Prior to version 1.0.29, the tool relay authorizes requests from any process whose command line contains an allowed helper script path (the Codex CLI shim). This allows untrusted code executing in the sandbox to invoke arbitrary host-exposed tools, including secret lookups, deployment operations, and cloud API calls without a corresponding model-authorized tool-call event. Exploitation requires a Linux environment (the vulnerable fallback checks `process.platform === 'linux'` and reads `/proc`); an active harness session with one or more host-provided tools; and untrusted code executing in the sandbox (e.g. a malicious dependency, build script, or lifecycle hook) The fix in version 1.0.29 removes the process-path authorization fallback entirely. Relay requests are now only accepted after exact, short-lived, one-time authorization matching the tool name and input from a bridge-observed model event. Some workarounds are available. Do not run the Codex harness on untrusted repositories or with untrusted dependencies, and/or limit host-exposed tools to non-sensitive operations when working with untrusted code. | 24 Oct |
| CVE-2026-64651 | 6.3 medium | 0.00% | no CPE yet | The `@ai-sdk/harness-opencode` tool connects HarnessAgent to OpenCode through a sandboxed bridge. Prior to version 1.0.28, the tool relay authorizes requests from any process whose command line contains an allowed helper script path (`host-tool-mcp.mjs`). This allows untrusted code executing in the sandbox to invoke arbitrary host-exposed tools including secret lookups, deployment operations, and cloud API calls without a corresponding model-authorized tool-call event. Exploitation requires a Linux environment (the vulnerable fallback checks `process.platform === 'linux'` and reads `/proc`); an active harness session with one or more host-provided tools; and untrusted code executing in the sandbox (e.g. a malicious dependency, build script, or lifecycle hook) The fix in version 1.0.28 removes the process-path authorization fallback entirely. Relay requests are now only accepted after exact, short-lived, one-time authorization matching the tool name and input from a bridge-observed model event. Some workarounds are available. Do not run the OpenCode harness on untrusted repositories or with untrusted dependencies, and/or limit host-exposed tools to non-sensitive operations when working with untrusted code. | 24 Oct |
| CVE-2026-114386 | 6.3 medium | 0.00% | no CPE yet | Pydantic AI Harness before 0.59.0 fails to enforce UsageLimits.tool_calls_limit on nested tool calls dispatched from run_code in the CodeMode toolset. Users or prompt-injected content can make the sandboxed program invoke tools via run_nested_call() without check_before_tool_call, exceeding operator limits on costly or side-effecting tools. | 22 Oct |
| CVE-2026-53656 | 6.3 medium | 0.00% | no CPE yet | FiftyOne is an open-source platform for refining high-quality datasets and visual AI models. Prior to 1.17.0, the FiftyOne App/API server in fiftyone/server/app.py and the /media route in fiftyone/server/routes/media.py unconditionally return Access-Control-Allow-Origin: *. Because the embedded server is local and unauthenticated, a malicious website visited by the user can read cross-origin responses. The /media endpoint accepts a filesystem path, allowing a drive-by page to read files accessible to the server process and exfiltrate them without additional clicks. The allowed_origins configuration and FIFTYONE_ALLOWED_ORIGINS environment variable now make cross-origin access explicit, while the default policy is same-origin. This issue is fixed in version 1.17.0. | 12 Oct |
| CVE-2026-54785 | 6.2 medium | 0.00% | no CPE yet | gemini-bridge is a lightweight MCP server bridging AI agents to Google's Gemini AI via the official CLI. From 1.0.0 until 1.3.1, consult_gemini_with_files in inline mode read any file path supplied in the files argument without confining it to the working directory, then forwarded the contents to the Gemini CLI. Because the caller also controls query, the file contents are echoed back through the Gemini round-trip (and sent to Google), making this an arbitrary local file read. This issue is fixed in version 1.3.1. | 23 Oct |
| CVE-2026-57208 | 6.2 medium | 0.00% | no CPE yet | Entire CLI hooks into Git workflows to capture AI agent sessions. Prior to version 0.7.7, a path traversal vulnerability in Entire CLI allows an attacker with push access to the checkpoints repository to craft malicious checkpoint metadata that causes entire session resume or entire checkpoint rewind to write attacker-controlled transcript data outside of the expected session directory. The issue occurs because checkpoint metadata is fetched from the remote entire/checkpoints/v1 branch and the SessionID field was used to construct filesystem paths without validation in the restore path. A malicious SessionID containing absolute paths or path traversal sequences could cause arbitrary files on the victim’s machine to be overwritten. This issue has been patched in version 0.7.7. | 23 Oct |
| CVE-2026-100151 | 6.1 medium | 0.00% | kstover, ninja forms – contact form builder with calculators, quizzes, signatures & ai form builder | The Ninja Forms – Contact Form Builder with Calculators, Quizzes, Signatures & AI Form Builder plugin for WordPress is vulnerable to Reflected DOM-Based Cross-Site Scripting via 'querystring:<key>' URL Parameter in all versions up to, and including, 3.15.5 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. This bypass is triggered when an attacker supplies a matching field:<key> URL parameter that causes the server to fully resolve the {field:*} placeholder, forcing the client-side else-branch to abandon the server-sanitized value and reconstruct field content from the raw query string. | 13 h ago |
| CVE-2026-91152 | 6.1 medium | 0.00% | wpeverest, everest forms – contact form, payment form, quiz, survey & custom form builder with ai | The Everest Forms – Contact Form, Payment Form, Quiz, Survey & Custom Form Builder with AI plugin for WordPress is vulnerable to Reflected Cross-Site Scripting via the 'evf_toast_type' parameter in all versions up to, and including, 3.6.2 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. Exploitation requires the sibling evf_toast parameter to decode successfully via atob(), but this is a trivially attacker-controlled prerequisite. | Thu 29 Oct |
| CVE-2026-47250 | 6.1 medium | 0.00% | no CPE yet | mcp-server-kubernetes is a Model Context Protocol server for Kubernetes cluster management. Prior to version 3.7.0, the kubectl_generic tool in mcp-server-kubernetes passes user-supplied flags directly to kubectl without any allowlist, enabling a privilege escalation attack within Kubernetes environments. An attacker who already has limited cluster or codebase access, for example, a developer with pod-deployment permissions but not cluster-admin credentials, can plant a single structured JSON line in an application's log output. When an operator with a privileged kubeconfig uses the MCP server to read those logs and their AI agent follows the injected instruction, kubectl_generic is called with --server=https://attacker.example.com and --insecure-skip-tls-verify=true. kubectl sends all API requests, including the Authorization: Bearer <token> header from the operator's kubeconfig to the attacker's endpoint. The captured token can then be replayed directly against the real Kubernetes API server, granting the attacker the full RBAC permissions of the operator's service account. This issue has been patched in version 3.7.0. | Mon 26 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.