AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. In 1.16.1 and earlier, the manager role can store meta_page_title or meta_page_favicon through /api/admin/system-preferences, and MetaGenerator inserts those values into production homepage HTML without escaping attribute values or text content. The values pass unchanged through server/models/systemSettings.js and reach MetaGenerator.generate() in server/index.js. #assembleMeta() in server/utils/boot/MetaGenerator.js concatenates the stored values into HTML. When an administrator visits the homepage /, injected JavaScript can read the administrator JWT and use it to create API keys, access or modify workspace and chat data, delete users, and perform other administrator actions. server/endpoints/admin.js accepts the manager-controlled settings before server/models/systemSettings.js returns them unchanged. No fixed version is available as of this review.
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. From 1.11.1 until 1.14.1, userId/workspaceId scoping to the parsed-files read/delete paths was added. However, the POST /api/workspace/:slug/embed-parsed-file/:fileId flow still deletes the target file by primary key only, with no ownership check, inside two finally{} blocks that run even when the ownership-checked read fails. As a result a manager or admin (multi-user mode) can delete any other user's parsed file in any workspace — including workspaces they are not a member of — by enumerating integer fileIds. The server even returns "File not found" while still deleting the file. This vulnerability is fixed in 1.14.1.
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. Prior to 1.13.0, on Windows, the document folder listing route can accept an encoded absolute Windows path that resolves outside the intended documents directory. The shared path containment helper rejects POSIX-style "../" traversal but does not reject Windows-style parent paths returned by path.relative(), such as "..". This vulnerability is fixed in 1.13.0.
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. Prior to 1.13.0, an approved mobile device token created in single-user mode can survive single-user -> multi-user migration even when the device record has userId = null. In multi-user mode, that stale token is still accepted by the mobile authentication middleware. Because no user is attached to the request, downstream mobile handlers fall back to unscoped data-access branches and return workspaces and workspace content without per-user filtering. This permits a pre-migration mobile token to enumerate a workspace assigned only to another user and retrieve victim-owned thread metadata and chat content in multi-user mode. This vulnerability is fixed in 1.13.0.
CVE-2026-48116 is a command injection vulnerability in Mintplex-Labs anything-llm versions prior to 1.13.0. The issue arises because the filesystem-search-files agent skill passes an LLM-controlled pattern parameter to ripgrep without using an end-of-options separator, allowing specially crafted patterns to be interpreted as command options. This enables an attacker who can interact with the agent and has the filesystem plugin enabled to execute arbitrary commands inside the server container. The vulnerability is fixed in version 1.13.0.
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. Prior to 1.13.0, the AnythingLLM agent filesystem copy tool validates only the top-level source and destination paths. The recursive copy helper then descends into child entries using fs.stat() and copies files with fs.copyFile() without validating each child or rejecting symlinks. Because both APIs follow symlinks, a symlink nested inside an allowed source directory can point outside the allowed filesystem root and cause outside file contents to be copied into an allowed destination as a regular file. This vulnerability is fixed in 1.13.0.
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. Prior to version 1.12.1, GET /api/workspace/:slug/tts/:chatId in AnythingLLM returns the text-to-speech audio for another user's chat response within the same workspace because the route validates workspace membership but does not enforce ownership of the targeted chat row. As a result, an authenticated user can access another user's private assistant response in audio form if the chatId is known or guessed. This constitutes an insecure direct object reference (IDOR) affecting private chat response content exposed through the TTS endpoint. This issue has been patched in version 1.12.1.
CVE-2026-41318 is a stored cross-site scripting (XSS) vulnerability in Mintplex-Labs' AnythingLLM application versions prior to 1.12.1. The issue arises from unsafe handling of markdown image alt text in the in-chat markdown renderer, specifically in the Chartable component, which does not sanitize chart captions. An attacker able to influence the LLM's output can inject malicious scripts that execute in other users' browsers when they view the affected chat. Version 1.12.1 includes a patch addressing this vulnerability.
CVE-2026-5627 is a critical path traversal vulnerability in mintplex-labs/anything-llm up to version 1.9.1, specifically in the AgentFlows component. It allows attackers with high privileges to bypass directory restrictions via crafted input to loadFlow and deleteFlow methods, potentially accessing or deleting arbitrary JSON files on the server. This can lead to disclosure of sensitive configuration files containing API keys or denial of service by deleting important files like package.json. The vulnerability is fixed in version 1.12.1. No official patch link or advisory is provided in the data, and no known exploits are reported in the wild.
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. In 1.11.1 and earlier, The ImportedPlugin.importCommunityItemFromUrl() function in server/utils/agents/imported.js downloads a ZIP file from a community hub URL and extracts it using AdmZip.extractAllTo() without validating file paths within the archive. This enables a Zip Slip path traversal attack that can lead to arbitrary code execution.
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. In 1.11.1 and earlier, in multi-user mode, AnythingLLM blocks suspended users on the normal JWT-backed session path, but it does not block them on the browser extension API key path. If a user already has a valid brx-... browser extension API key, that key continues to work after suspension. As a result, a suspended user can still access browser extension endpoints, read reachable workspace metadata, and continue upload or embed operations even though normal authenticated requests are rejected.
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. In 1.11.1 and earlier, The two generic system-preferences endpoints allow manager role access, while every other surface that touches the same settings is restricted to admin only. Because of this inconsistency, a manager can call the generic endpoints directly to read plaintext SQL database credentials and overwrite admin-only global settings such as the default system prompt and the Community Hub API key.
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. In 1.11.1 and earlier, a SQL injection vulnerability in the built-in SQL Agent plugin allows any user who can invoke the agent to execute arbitrary SQL commands on connected databases. The getTableSchemaSql() method in all three database connectors (MySQL, PostgreSQL, MSSQL) constructs SQL queries using direct string concatenation of the table_name parameter without sanitization or parameterization.
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. In 1.11.1 and earlier, AnythingLLM Desktop contains a Streaming Phase XSS vulnerability in the chat rendering pipeline that escalates to Remote Code Execution on the host OS due to insecure Electron configuration. This works with default settings and requires no user interaction beyond normal chat usage. The custom markdown-it image renderer in frontend/src/utils/chat/markdown.js interpolates token.content directly into the alt attribute without HTML entity escaping. The PromptReply component renders this output via dangerouslySetInnerHTML without DOMPurify sanitization — unlike HistoricalMessage which correctly applies DOMPurify.sanitize().
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. In 1.11.1 and earlier, On default installations where no password or API key has been configured, all HTTP endpoints and the agent WebSocket lack authentication, and the server's CORS policy accepts any origin. AnythingLLM Desktop binds to 127.0.0.1 (loopback) by default. Modern browsers (Chrome, Edge, Firefox) implement Private Network Access (PNA). This explicitly blocks public websites from making requests to local IP addresses. Exploitation is only viable from within the same local network (LAN) due to browser-level blocking of public-to-private requests.
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. Prior to version 1.10.0, a critical Path Traversal vulnerability in the DrupalWiki integration allows a malicious admin (or an attacker who can convince an admin to configure a malicious DrupalWiki URL) to write arbitrary files to the server. This can lead to Remote Code Execution (RCE) by overwriting configuration files or writing executable scripts. Version 1.10.0 fixes the issue.
AnythingLLM is an application that turns pieces of content into context that any LLM can use as references during chatting. If AnythingLLM prior to version 1.10.0 is configured to use Qdrant as the vector database with an API key, this QdrantApiKey could be exposed in plain text to unauthenticated users via the `/api/setup-complete` endpoint. Leakage of QdrantApiKey allows an unauthenticated attacker full read/write access to the Qdrant vector database instance used by AnythingLLM. Since Qdrant often stores the core knowledge base for RAG in AnythingLLM, this can lead to complete compromise of the semantic search / retrieval functionality and indirect leakage of confidential uploaded documents. Version 1.10.0 patches the issue.
An authentication bypass vulnerability exists in AnythingLLM v1.8.5 in via the /api/workspaces endpoint. The endpoint fails to implement proper authentication checks, allowing unauthenticated remote attackers to enumerate and retrieve detailed information about all configured workspaces. Exposed data includes: workspace identifiers (id, name, slug), AI model configurations (chatProvider, chatModel, agentProvider), system prompts (openAiPrompt), operational parameters (temperature, history length, similarity thresholds), vector search settings, chat modes, and timestamps.
In version 1.5.5 of mintplex-labs/anything-llm, the `/setup-complete` API endpoint allows unauthorized users to access sensitive system settings. The data returned by the `currentSettings` function includes sensitive information such as API keys for search engines, which can be exploited by attackers to steal these keys and cause loss of user assets.
A vulnerability in mintplex-labs/anything-llm prior to version 1.2.2 allows for Prisma injection. The issue exists in the API endpoint "/embed/:embedId/stream-chat" where user-provided JSON is directly taken to the Prisma library's where clause. An attacker can exploit this by providing a specially crafted JSON object, such as {"sessionId":{"not":"a"}}, causing Prisma to return all data from the table. This can lead to unauthorized access to all user queries in embedded chat mode.