BerriAI/litellm · error · HTTPException

Rejected message. This is a prompt injection attack.

Error message

Rejected message. This is a prompt injection attack.

What it means

Error "Rejected message. This is a prompt injection attack." thrown in BerriAI/litellm.

Source

Thrown at litellm/proxy/hooks/prompt_injection_detection.py:170

                    "image_generation",
                    "moderation",
                    "audio_transcription",
                ]
            except Exception:
                self.print_verbose(
                    f"Call Type - {call_type}, not in accepted list - ['completion','embeddings','image_generation','moderation','audio_transcription']"
                )
                return data
            formatted_prompt: Final = get_formatted_prompt(data=data, call_type=call_type)

            is_prompt_attack = False

            if self.prompt_injection_params is not None:
                # 1. check if heuristics check turned on
                if self.prompt_injection_params.heuristics_check is True:
                    is_prompt_attack = self.check_user_input_similarity(user_input=formatted_prompt)
                    if is_prompt_attack is True:
                        raise HTTPException(
                            status_code=400,
                            detail={"error": "Rejected message. This is a prompt injection attack."},
                        )
                # 2. check if vector db similarity check turned on [TODO] Not Implemented yet
                if self.prompt_injection_params.vector_db_check is True:
                    pass
            else:
                is_prompt_attack = self.check_user_input_similarity(user_input=formatted_prompt)

            if is_prompt_attack is True:
                raise HTTPException(
                    status_code=400,
                    detail={"error": "Rejected message. This is a prompt injection attack."},
                )

            return data

        except HTTPException as e:

View on GitHub (pinned to 77b7c6c40c)

Solutions

1. Remove prompt-injection content (e.g. instructions to ignore prior directives) from the message.
2. If this is a false positive, review the detection model and threshold configuration.

When it happens

Trigger: Thrown at litellm/proxy/hooks/prompt_injection_detection.py:170 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/2a6bcda5ea76dbb1. Report an issue: GitHub.