MaxKB is an open-source AI assistant for enterprise. Prior to version 2.10.6-lts, function-library code running under the LDPRELOAD sandbox can invoke ctypes.CDLL from an importlib.abc.MetaPathFinder callback so the dlopen call-stack heuristic sees a Python import frame, then use unhooked dlsym with RTLDNEXT to resolve glibc's real syscall and bypass the sandbox syscall blacklist. An authenticated workspace member can consequently read or write files, execute processes, or access networks as the sandbox user. This issue is fixed in version 2.10.6-lts.

MaxKB is an open-source AI assistant for enterprise. From version 2.0.0 through 2.10.2-lts, document and paragraph operate routes authorize only knowledgeid in the request path, then query the target Document by documentid or Paragraph by paragraphid without confirming that the object belongs to that knowledge base. A normal workspace user with a known victim document or paragraph UUID can use an attacker-owned knowledge-base path to read or modify content in another user's knowledge base. No fixed version is available as of this review.

MaxKB is an open-source AI assistant for enterprise. From version 2.0.0 through 2.9.2, a lowest-role workspace member denied access to a tool by WorkspaceUserResourcePermission can still bind its identifier through toolids, skilltoolids, or mcptoolids and execute it through the agent or workflow dispatch path. The dispatch path does not reapply the per-tool grant enforced by dedicated tool routes, and tool execution decrypts server-side initparams, allowing the caller to receive credentials carried by the denied tool. No fixed version is available as of this review.

MaxKB is an open-source AI assistant for enterprise. In 2.10.2-lts and earlier, management chat-record routes authorize the path applicationid but load records using global chatid and chatrecordid values without confirming that the chat belongs to the authorized application. A normal user in the same workspace who knows the chatid of a persisted non-debug record for a victim's published application can place it under a separate attacker-owned application path to read victim chat records. An attacker with an owned knowledge base and document can also use addknowledge to copy victim answers while updating the victim record's improveparagraphidlist. No fixed version is available as of this review.

MaxKB is an open-source AI assistant for enterprise. In 2.10.2-lts and earlier, a normal workspace user who knows another user's active MCP toolid in the same workspace can retrieve the hidden tool through the tool-detail route because it does not apply the per-resource authorization used by the list route. The response includes Tool.code, which may contain MCP server configuration and headers, and an attacker who can create or edit an attacker-owned workflow can place the same foreign mcptoolid in a workflow mcp-node so workflow debug uses the owner's MCP configuration without verifying permission to use that tool. No fixed version is available as of this review.

MaxKB is an open-source AI assistant for enterprise. In 2.10.2-lts and earlier, a normal user in the same workspace can obtain another user's applicationid from the homepage application question-ranking endpoint when the published victim application has ranking activity in the selected date range and the attacker knows or guesses its name, even though direct application detail and debug-open routes deny access. An attacker who can create and publish a workflow application can place the disclosed identifier in an attacker-owned workflow application-node, trigger that workflow, receive output generated by the victim application, and create durable applicationchat and applicationchatrecord rows under the victim application because save and runtime paths do not verify permission to use the referenced application. No fixed version is available as of this review.

MaxKB is an open-source AI assistant for enterprise. In 2.10.2-lts and earlier, the /chat/api/mcp authentication path looks up an ApplicationApiKey using only its secret and active status, without enforcing the ispermanent and expiretime checks used by the standard chat API. A previously valid, non-permanent application key that remains active after expiration can initialize the MCP endpoint, list the published application's tool metadata, invoke tools/call, and create durable applicationchat and applicationchatrecord state even though /chat/api/open rejects the same key. No fixed version is available as of this review.

MaxKB is an open-source AI assistant for enterprise. In versions 2.0.0 through 2.10.4-lts, UpdateStoreTool.updatetool passes caller-supplied downloadurl and downloadcallbackurl values to requests.get without equivalent trusted-host and redirect validation, allowing an authenticated workspace user to make the server request internal, loopback, link-local, or cloud metadata URLs. A fix is present on the v2 branch but has not yet been included in a published release.

MaxKB is an open-source AI assistant for enterprise. MaxKB 2.8.0 and prior are vulnerable to a server-side request forgery (SSRF) bypass in the OSS file service URL fetch functionality due to inconsistent DNS resolution between validation and actual request execution, allowing attackers to access internal network services. This vulnerability is fixed in 2.8.1.

MaxKB is an open-source AI assistant for enterprise. MaxKB 2.8.0 and prior are vulnerable to a broken access control vulnerability in the OSS file service URL fetch API (chat/api/oss/geturl). The endpoint uses applicationid from the URL path without validating ownership, allowing attackers to perform operations under other applications’ policies. This vulnerability is fixed in 2.8.1.

MaxKB is an open-source AI assistant for enterprise. Prior to 2.9.0, MaxKB's webhook trigger endpoint (/api/trigger/v1/webhook/{triggerid}) is accessible without authentication. The WebhookAuth class unconditionally returns (None, {}), which Django REST Framework interprets as successful authentication. Combined with optional per-trigger token verification and no backend enforcement of token requirements, any unauthenticated attacker who knows a valid trigger ID can invoke webhook triggers to execute their bound tasks. This vulnerability is fixed in 2.9.0.

MaxKB is an open-source AI assistant for enterprise. Prior to 2.9.1, SSRF via workflowtemplate Import. Authenticated users can supply arbitrary URLs in workflowtemplate.downloadUrl which are fetched server-side without any URL validation or internal IP filtering. This vulnerability is fixed in 2.9.1.

MaxKB is an open-source AI assistant for enterprise. Prior to 2.9.1, user passwords are stored using unsalted MD5 hashes, making them trivially crackable via rainbow tables or GPU-accelerated brute force (hashcat). This vulnerability is fixed in 2.9.1.

MaxKB is an open-source AI assistant for enterprise. Prior to 2.8.1, MaxKB v2.8.0 and prior are vulnerable to a server-side request forgery (SSRF) bypass in the OSS file service URL fetch (chat/api/oss/geturl) endpoint. The vulnerability exists due to inconsistent URL parsing between the urlparse validation function and the requests HTTP client, allowing attackers to access internal network services. This vulnerability is fixed in 2.8.1.

MaxKB is an open-source AI assistant for enterprise. Versions 2.7.1 and below contain a Stored Cross-Site Scripting (XSS) vulnerability where the frontend's MdRenderer.vue component parses custom <iframerender> tags from LLM responses or Application Prologue configurations, bypassing standard Markdown sanitization and XSS filtering. The unsanitized HTML content is passed to the IframeRender.vue component, which renders it directly into an <iframe> via the srcdoc attribute configured with sandbox="allow-scripts allow-same-origin". This can be a dangerous combination, allowing injected scripts to escape the iframe and execute JavaScript in the parent window using window.parent. Since the Prologue is rendered for any user visiting an application's chat interface, this results in a high-impact Stored XSS that can lead to session hijacking, unauthorized actions, and sensitive data exposure. This issue has been fixed in version 2.8.0.

MaxKB is an open-source AI assistant for enterprise. Versions 2.7.1 and below contain a Stored Cross-Site Scripting (XSS) vulnerability that allows authenticated users to inject arbitrary HTML and JavaScript into the Application prologue (Opening Remarks) field by wrapping malicious payloads in <htmlrander> tags. The backend fails to sanitize or encode HTML entities in the prologue field when applications are created or updated via the /admin/api/workspace/{workspaceid}/application endpoint, storing the raw payload directly in the database. The frontend then renders this content using an innerHTML-equivalent mechanism, trusting <htmlrander>-wrapped content to be safe, which enables persistent DOM-based Stored XSS execution against any visitor who opens the affected chatbot interface. Exploitation can lead to session hijacking, unauthorized actions performed on behalf of victims (such as deleting workspaces or applications), and sensitive data exposure. This issue has been fixed in version 2.8.0.

MaxKB is an open-source AI assistant for enterprise. In versions 2.7.1 and below, an authenticated user can bypass sandbox result validation and spoof tool execution results by exploiting Python frame introspection to read the wrapper's UUID from its bytecode constants, then writing a forged result directly to file descriptor 1 (bypassing stdout redirection). By calling sys.exit(0), the attacker terminates the wrapper before it prints the legitimate output, causing the MaxKB service to parse and trust the spoofed response as the genuine tool result. This issue has been fixed in version 2.8.0.

MaxKB is an open-source AI assistant for enterprise. In versions 2.7.1 and below, the chat export feature is vulnerable to Improper Neutralization of Formula Elements in a CSV File. When an administrator exports the application chat history to an Excel file (.xlsx) via the /admin/api/workspace/{workspaceid}/application/{applicationid}/chat/export endpoint, strings starting with formula characters are written directly without proper sanitization. Opening this file in spreadsheet applications like Microsoft Excel can lead to Arbitrary Code Execution (RCE) on the administrator's workstation via Dynamic Data Exchange (DDE). The issue is a variant of CVE-2025-4546, which fixed the exact same pattern in apps/dataset/serializers/documentserializers.py but missed the application chat export sink. This issue has been fixed in version 2.8.0.

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.

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/{accesstoken}), 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.

MaxKB is an open-source AI assistant for enterprise. Versions 2.7.1 and below contain a sandbox escape vulnerability in the ToolExecutor component. By leveraging Python's ctypes library to execute raw system calls, an authenticated attacker with workspace privileges can bypass the LDPRELOAD-based sandbox.so module to achieve arbitrary code execution via direct kernel system calls, enabling full network exfiltration and container compromise. The library intercepts critical standard system functions such as execve, system, connect, and open. It also intercepts mprotect to prevent PROTEXEC (executable memory) allocations within the sandboxed Python processes, but pkeymprotect is not blocked. This issue has been fixed in version 2.8.0.

MaxKB is an open-source AI assistant for enterprise. In versions 2.7.1 and below, an incomplete sandbox protection mechanism allows an authenticated user with tool execution privileges to escape the LDPRELOAD-based sandbox. By env command the attacker can clear the environment variables and drop the sandbox.so hook, leading to unrestricted Remote Code Execution (RCE) and network access. MaxKB restricts untrusted Python code execution via the Tool Debug API by injecting sandbox.so through the LDPRELOAD environment variable. This intercepts sensitive C library functions (like execve, socket, open) to restrict network and file access. However, a patch allowed the /usr/bin/env utility to be executed by the sandboxed user. When an attacker is permitted to create subprocesses, they can execute the env -i python command. The -i flag instructs env to completely clear all environment variables before running the target program. This effectively drops the LDPRELOAD environment variable. The newly spawned Python process will therefore execute natively without any sandbox hooks, bypassing all network and file system restrictions. This issue has been fixed in version 2.8.0.

MaxKB is an open-source AI assistant for enterprise. In versions 2.7.1 and below, sandbox network protection can be bypassed by using socket.sendto() with the MSGFASTOPEN flag. This allows authenticated user with tool-editing permissions to reach internal services that are explicitly blocked by the sandbox's banned hosts configuration. MaxKB's sandbox uses LDPRELOAD to hook the connect() function and block connections to banned IPs, but Linux's sendto() with the MSGFASTOPEN flag can establish TCP connections directly through the kernel without ever calling connect(), completely bypassing the IP validation. Although sendto is listed in the syscall() wrapper, this is ineffective because glibc invokes the kernel syscall directly rather than routing through the hooked syscall() function. This issue has been fixed in version 2.8.0.

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 mcpservers directly from user-supplied JSON remains completely unpatched. Since mcpsource 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.

MaxKB is an open-source AI assistant for enterprise. Versions 2.3.1 and below have improper file permissions which allow attackers to overwrite the built-in dynamic linker and other critical files, potentially resulting in privilege escalation. This issue is fixed in version 2.4.0.

MaxKB is an open-source AI assistant for enterprise. In versions 2.3.1 and below, the tool module allows an attacker to escape the sandbox environment and escalate privileges under certain concurrent conditions. This issue is fixed in version 2.4.0.

MaxKB is an open-source AI assistant for enterprise. In versions prior to 2.3.1, a user can get sensitive informations by Python code in tool module, although the process run in sandbox. Version 2.3.1 fixes the issue.

MaxKB is an open-source AI assistant for enterprise. In versions prior to 2.3.1, a user can access internal network services such as databases through Python code in the tool module, although the process runs in a sandbox. Version 2.3.1 fixes the issue.

MaxKB is an open-source AI assistant for enterprise. Prior to versions 1.10.9-lts and 2.0.0, a Remote Command Execution vulnerability exists in the MCP call. Versions 1.10.9-lts and 2.0.0 fix the issue.

MaxKB is an open-source AI assistant for enterprise. Prior to version 2.0.0, the sandbox design rules can be bypassed because MaxKB only restricts the execution permissions of files in a specific directory. Therefore, an attacker can use the shutil.copy2 method in Python to copy the command they want to execute to the executable directory. This bypasses directory restrictions and reverse shell. Version 2.0.0 fixes the issue.