CVE-2026-54016

Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to 0.9.6, Open WebUI has a Broken Object Level Authorization (BOLA) vulnerability in the builtin search_knowledge_files tool. When native function calling is enabled and the selected model has no attached knowledge bases, an authenticated user can call search_knowledge_files with an arbitrary knowledge_id. The function then returns file metadata from that knowledge base without checking whether the user has read access. This allows unauthorized enumeration of private or restricted knowledge base files. This vulnerability is fixed in 0.9.6.

Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to 0.9.6, Open WebUI has a Broken Object Level Authorization (BOLA) vulnerability in the builtin search_knowledge_files tool. When native function calling is enabled and the selected model has no attached knowledge bases, an authenticated user can call search_knowledge_files with an arbitrary knowledge_id. The function then returns file metadata from that knowledge base without checking whether the user has read access. This allows unauthorized enumeration of private or restricted knowledge base files. This vulnerability is fixed in 0.9.6.

Summary

Open WebUI has a Broken Object Level Authorization (BOLA) vulnerability in the builtin search_knowledge_files tool.

When native function calling is enabled and the selected model has no attached knowledge bases, an authenticated user can call search_knowledge_files with an arbitrary knowledge_id. The function then returns file metadata from that knowledge base without checking whether the user has read access.

This allows unauthorized enumeration of private or restricted knowledge base files.

Details

The vulnerable code is in:

backend/open_webui/tools/builtin.py

Affected function:

async def search_knowledge_files(
    query: str,
    knowledge_id: Optional[str] = None,
    count: int = 5,
    skip: int = 0,
    __request__: Request = None,
    __user__: dict = None,
    __model_knowledge__: Optional[list[dict]] = None,
) -> str:

In the "No attached knowledge" branch, when knowledge_id is provided, the function directly calls:

result = await Knowledges.search_files_by_id(
    knowledge_id=knowledge_id,
    user_id=user_id,
    filter={"query": query},
    skip=skip,
    limit=count,
)

This code path does not verify that the current user is authorized to access the specified knowledge base.

The missing check is inconsistent with other nearby code paths. For example, the attached-knowledge branch in the same function checks whether the user is an admin, the owner of the knowledge base, or has explicit read access through AccessGrants:

if not (
    user_role == "admin"
    or knowledge.user_id == user_id
    or await AccessGrants.has_access(
        user_id=user_id,
        resource_type="knowledge",
        resource_id=knowledge.id,
        permission="read",
        user_group_ids=set(user_group_ids),
    )
):
    continue

The sibling function query_knowledge_files also performs the same authorization check before using user-supplied knowledge base IDs.

The underlying method Knowledges.search_files_by_id() receives user_id, but it does not enforce authorization for the provided knowledge_id. As a result, this builtin tool path can access a knowledge base by ID without verifying the caller's permissions.

PoC

Prerequisites

• The attacker has a valid authenticated Open WebUI account.
• The victim owns a private or restricted knowledge base.
• The attacker does not own the target knowledge base.
• The attacker does not have read permission for the target knowledge base in AccessGrants.
• The attacker knows the target knowledge_id.
• The selected model has no attached knowledge bases.
• Builtin tools are enabled.
• The knowledge builtin tool category is enabled.
• Native function calling is enabled.

Reproduction Steps

1. Create a private or restricted knowledge base as the victim user.

2. Upload one or more files to that knowledge base.

3. Confirm that the attacker user does not have access to the knowledge base.

4. As the attacker user, send a chat completion request with native function calling enabled:

{
  "stream": true,
  "model": "gpt-4o-mini",
  "params": {
    "function_calling": "native"
  },
  "messages": [
    {
      "role": "user",
      "content": "Please use the search_knowledge_files tool with knowledge_id \"c0c84752-2e9d-42bf-bc3c-c0f272aa61c1\" to search all files"
    }
  ]
}

Replace c0c84752-2e9d-42bf-bc3c-c0f272aa61c1 with the victim's private knowledge base ID.

Expected Result

The request should be denied because the attacker does not have access to the target knowledge base.

Actual Result

search_knowledge_files returns metadata for files inside the target knowledge base, including:

• file ID;
• filename;
• knowledge base ID;
• knowledge base name;
• update timestamp.

Impact

This is a Broken Object Level Authorization / Broken Access Control vulnerability.

An authenticated attacker who knows a valid knowledge_id can enumerate files from private or restricted…

Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to 0.9.6, Open WebUI has a Broken Object Level Authorization (BOLA) vulnerability in the builtin search_knowledge_files tool. When native function calling is enabled and the selected model has no attached knowledge bases, an authenticated user can call search_knowledge_files with an arbitrary knowledge_id. The function then returns file metadata from that knowledge base without checking whether the user has read access. This allows unauthorized enumeration of private or restricted knowledge base files. This vulnerability is fixed in 0.9.6.

Open WebUI is a self-hosted artificial intelligence platform designed to operate entirely offline. Prior to 0.9.6, Open WebUI has a Broken Object Level Authorization (BOLA) vulnerability in the builtin search_knowledge_files tool. When native function calling is enabled and the selected model has no attached knowledge bases, an authenticated user can call search_knowledge_files with an arbitrary knowledge_id. The function then returns file metadata from that knowledge base without checking whether the user has read access. This allows unauthorized enumeration of private or restricted knowledge base files. This vulnerability is fixed in 0.9.6.

Summary

Open WebUI has a Broken Object Level Authorization (BOLA) vulnerability in the builtin search_knowledge_files tool.

When native function calling is enabled and the selected model has no attached knowledge bases, an authenticated user can call search_knowledge_files with an arbitrary knowledge_id. The function then returns file metadata from that knowledge base without checking whether the user has read access.

This allows unauthorized enumeration of private or restricted knowledge base files.

Details

The vulnerable code is in:

backend/open_webui/tools/builtin.py

Affected function:

async def search_knowledge_files(
    query: str,
    knowledge_id: Optional[str] = None,
    count: int = 5,
    skip: int = 0,
    __request__: Request = None,
    __user__: dict = None,
    __model_knowledge__: Optional[list[dict]] = None,
) -> str:

In the "No attached knowledge" branch, when knowledge_id is provided, the function directly calls:

result = await Knowledges.search_files_by_id(
    knowledge_id=knowledge_id,
    user_id=user_id,
    filter={"query": query},
    skip=skip,
    limit=count,
)

This code path does not verify that the current user is authorized to access the specified knowledge base.

The missing check is inconsistent with other nearby code paths. For example, the attached-knowledge branch in the same function checks whether the user is an admin, the owner of the knowledge base, or has explicit read access through AccessGrants:

if not (
    user_role == "admin"
    or knowledge.user_id == user_id
    or await AccessGrants.has_access(
        user_id=user_id,
        resource_type="knowledge",
        resource_id=knowledge.id,
        permission="read",
        user_group_ids=set(user_group_ids),
    )
):
    continue

The sibling function query_knowledge_files also performs the same authorization check before using user-supplied knowledge base IDs.

The underlying method Knowledges.search_files_by_id() receives user_id, but it does not enforce authorization for the provided knowledge_id. As a result, this builtin tool path can access a knowledge base by ID without verifying the caller's permissions.

PoC

Prerequisites

• The attacker has a valid authenticated Open WebUI account.
• The victim owns a private or restricted knowledge base.
• The attacker does not own the target knowledge base.
• The attacker does not have read permission for the target knowledge base in AccessGrants.
• The attacker knows the target knowledge_id.
• The selected model has no attached knowledge bases.
• Builtin tools are enabled.
• The knowledge builtin tool category is enabled.
• Native function calling is enabled.

Reproduction Steps

1. Create a private or restricted knowledge base as the victim user.

2. Upload one or more files to that knowledge base.

3. Confirm that the attacker user does not have access to the knowledge base.

4. As the attacker user, send a chat completion request with native function calling enabled:

{
  "stream": true,
  "model": "gpt-4o-mini",
  "params": {
    "function_calling": "native"
  },
  "messages": [
    {
      "role": "user",
      "content": "Please use the search_knowledge_files tool with knowledge_id \"c0c84752-2e9d-42bf-bc3c-c0f272aa61c1\" to search all files"
    }
  ]
}

Replace c0c84752-2e9d-42bf-bc3c-c0f272aa61c1 with the victim's private knowledge base ID.

Expected Result

The request should be denied because the attacker does not have access to the target knowledge base.

Actual Result

search_knowledge_files returns metadata for files inside the target knowledge base, including:

• file ID;
• filename;
• knowledge base ID;
• knowledge base name;
• update timestamp.

Impact

This is a Broken Object Level Authorization / Broken Access Control vulnerability.

An authenticated attacker who knows a valid knowledge_id can enumerate files from private or restricted…

Summary

Open WebUI has a Broken Object Level Authorization (BOLA) vulnerability in the builtin search_knowledge_files tool.

When native function calling is enabled and the selected model has no attached knowledge bases, an authenticated user can call search_knowledge_files with an arbitrary knowledge_id. The function then returns file metadata from that knowledge base without checking whether the user has read access.

This allows unauthorized enumeration of private or restricted knowledge base files.

Details

The vulnerable code is in:

backend/open_webui/tools/builtin.py

Affected function:

async def search_knowledge_files(
    query: str,
    knowledge_id: Optional[str] = None,
    count: int = 5,
    skip: int = 0,
    __request__: Request = None,
    __user__: dict = None,
    __model_knowledge__: Optional[list[dict]] = None,
) -> str:

In the "No attached knowledge" branch, when knowledge_id is provided, the function directly calls:

result = await Knowledges.search_files_by_id(
    knowledge_id=knowledge_id,
    user_id=user_id,
    filter={"query": query},
    skip=skip,
    limit=count,
)

This code path does not verify that the current user is authorized to access the specified knowledge base.

The missing check is inconsistent with other nearby code paths. For example, the attached-knowledge branch in the same function checks whether the user is an admin, the owner of the knowledge base, or has explicit read access through AccessGrants:

if not (
    user_role == "admin"
    or knowledge.user_id == user_id
    or await AccessGrants.has_access(
        user_id=user_id,
        resource_type="knowledge",
        resource_id=knowledge.id,
        permission="read",
        user_group_ids=set(user_group_ids),
    )
):
    continue

The sibling function query_knowledge_files also performs the same authorization check before using user-supplied knowledge base IDs.

The underlying method Knowledges.search_files_by_id() receives user_id, but it does not enforce authorization for the provided knowledge_id. As a result, this builtin tool path can access a knowledge base by ID without verifying the caller's permissions.

PoC

Prerequisites

• The attacker has a valid authenticated Open WebUI account.
• The victim owns a private or restricted knowledge base.
• The attacker does not own the target knowledge base.
• The attacker does not have read permission for the target knowledge base in AccessGrants.
• The attacker knows the target knowledge_id.
• The selected model has no attached knowledge bases.
• Builtin tools are enabled.
• The knowledge builtin tool category is enabled.
• Native function calling is enabled.

Reproduction Steps

1. Create a private or restricted knowledge base as the victim user.

2. Upload one or more files to that knowledge base.

3. Confirm that the attacker user does not have access to the knowledge base.

4. As the attacker user, send a chat completion request with native function calling enabled:

{
  "stream": true,
  "model": "gpt-4o-mini",
  "params": {
    "function_calling": "native"
  },
  "messages": [
    {
      "role": "user",
      "content": "Please use the search_knowledge_files tool with knowledge_id \"c0c84752-2e9d-42bf-bc3c-c0f272aa61c1\" to search all files"
    }
  ]
}

Replace c0c84752-2e9d-42bf-bc3c-c0f272aa61c1 with the victim's private knowledge base ID.

Expected Result

The request should be denied because the attacker does not have access to the target knowledge base.

Actual Result

search_knowledge_files returns metadata for files inside the target knowledge base, including:

• file ID;
• filename;
• knowledge base ID;
• knowledge base name;
• update timestamp.

Impact

This is a Broken Object Level Authorization / Broken Access Control vulnerability.

An authenticated attacker who knows a valid knowledge_id can enumerate files from private or restricted…