Gemini Safety Filtering

When using Gemini AI for translation or speech recognition tasks, you may sometimes encounter errors such as "response content flagged."

This occurs because Gemini imposes safety restrictions on the content it processes. Although the code allows for some adjustments and uses the most lenient setting of "Block None," the final decision to filter content is determined by Gemini's comprehensive evaluation.

The adjustable safety filters in the Gemini API cover the following categories. Content outside these categories cannot be adjusted via code:

Category	Description
Harassment	Negative or harmful comments targeting identity and/or protected attributes.
Hate Speech	Rude, disrespectful, or profane content.
Sexually Explicit	References to sexual acts or other obscene content.
Dangerous Content	Content that promotes, facilitates, or encourages harmful behavior.
Civic Integrity	Queries related to elections.

The table below explains the blocking settings that can be applied to each category in the code.

For example, if you set the blocking threshold for the Hate Speech category to Block Only High, the system will block all parts with a high probability of containing hate speech but allow any parts with a low probability of containing dangerous content.

Threshold (Google AI Studio)	Threshold (API)	Description
Block None	BLOCK_NONE	Always display content, regardless of the likelihood of unsafe content.
Block Only High	BLOCK_ONLY_HIGH	Block content when the probability of unsafe content is high.
Block Medium and Above	BLOCK_MEDIUM_AND_ABOVE	Block content when the likelihood of unsafe content is medium or high.
Block Low and Above	BLOCK_LOW_AND_ABOVE	Block content when the likelihood of unsafe content is low, medium, or high.
Not Applicable	HARM_BLOCK_THRESHOLD_UNSPECIFIED	Threshold unspecified; default threshold is used for blocking.

You can enable BLOCK_NONE in the code with the following settings:

safetySettings = [
    {
        "category": HarmCategory.HARM_CATEGORY_HARASSMENT,
        "threshold": HarmBlockThreshold.BLOCK_NONE,
    },
    {
        "category": HarmCategory.HARM_CATEGORY_HATE_SPEECH,
        "threshold": HarmBlockThreshold.BLOCK_NONE,
    },
    {
        "category": HarmCategory.HARM_CATEGORY_SEXUALLY_EXPLICIT,
        "threshold": HarmBlockThreshold.BLOCK_NONE,
    },
    {
        "category": HarmCategory.HARM_CATEGORY_DANGEROUS_CONTENT,
        "threshold": HarmBlockThreshold.BLOCK_NONE,
    },
]

model = genai.GenerativeModel('gemini-2.0-flash-exp')
model.generate_content(
                message,
                safety_settings=safetySettings
)

However, note that even if all settings are set to BLOCK_NONE, Gemini may still filter content based on contextual safety evaluations.

How to Reduce the Likelihood of Safety Restrictions?

Generally, the flash series models have more safety restrictions, while the pro and thinking series models are relatively less restrictive. Try switching between different models.

Additionally, when dealing with potentially sensitive content, sending smaller amounts of content at a time and reducing the context length can help lower the frequency of safety filtering.

How to Completely Disable Gemini's Safety Judgments and Allow All Content?

Link a foreign credit card and switch to a monthly-paid premium account.