BerriAI/litellm · warning · HTTPException

Keyword banned. Keyword={word}

Error message

Keyword banned. Keyword={word}

What it means

Raised as an HTTPException (400) by the BannedKeywords enterprise hook when a request's text (prompt, messages, etc., lowercased) contains any substring from banned_keywords_list. It is an intentional content-policy rejection, not a malfunction. The offending keyword is echoed back in the detail payload.

Source

Thrown at enterprise/enterprise_hooks/banned_keywords.py:62

                )
            except Exception as e:
                raise Exception(
                    f"An error occurred: {str(e)}, banned_keywords_list={banned_keywords_list}"
                )

    def print_verbose(self, print_statement, level: Literal["INFO", "DEBUG"] = "DEBUG"):
        if level == "INFO":
            verbose_proxy_logger.info(print_statement)
        elif level == "DEBUG":
            verbose_proxy_logger.debug(print_statement)

        if litellm.set_verbose is True:
            print(print_statement)  # noqa

    def test_violation(self, test_str: str):
        for word in self.banned_keywords_list:
            if word in test_str.lower():
                raise HTTPException(
                    status_code=400,
                    detail={"error": f"Keyword banned. Keyword={word}"},
                )

    async def async_pre_call_hook(
        self,
        user_api_key_dict: UserAPIKeyAuth,
        cache: DualCache,
        data: dict,
        call_type: str,  # "completion", "embeddings", "image_generation", "moderation"
    ):
        try:
            """
            - check if user id part of call
            - check if user id part of blocked list
            """
            self.print_verbose("Inside Banned Keyword List Pre-Call Hook")
            if is_text_content_call_type(call_type):

View on GitHub (pinned to 6c2dcb801b)

Solutions

1. Remove or rephrase the offending keyword occurrence in the request payload.
2. If this is a false positive, ask the proxy admin to remove or replace the keyword in banned_keywords_list.
3. Admins: prefer word-boundary matching or more specific keywords to avoid substring false positives.
4. Admins: check the proxy logs (print_verbose DEBUG output) to see which keyword matched.

from fastapi import HTTPException

try:
    resp = client.chat.completions.create(model="gpt-4o", messages=msgs, user="u1")
except HTTPException as e:
    if e.status_code == 400 and "Keyword banned" in str(e.detail):
        # content-policy rejection: surface to the end user, do not retry
        raise UserContentViolation(e.detail["error"]) from e
    raise

When it happens

Trigger: Any /chat/completions, /embeddings, or similar call routed through async_pre_call_hook where test_violation finds a banned word as a substring of the lowercased input text. Because matching is naive substring containment, a banned word also triggers when it appears inside a larger word.

Common situations: Users legitimately hitting a content filter configured by their organization; false positives from substring matching (e.g. banning 'assassin' but the text contains 'sassassin' style overlaps, or banned words appearing in code/logs being sent to the LLM).

Related errors

• User blocked from making LLM API Calls. User={user}
• Violated content safety policy. Category={category}
• Violated content safety policy
• Violated content safety policy
• Violated content safety policy

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/772d9ce010c35547. Report an issue: GitHub.