vercel/ai · error · Error

Amazon Bedrock request was moderated: ${reasons.join(', ')}

Error message

Amazon Bedrock request was moderated: ${reasons.join(', ')}

What it means

Nova Canvas can return a response whose status is 'Request Moderated', meaning Bedrock's content moderation blocked the prompt (and optionally images). The provider reads details['Moderation Reasons'] and throws with the joined reasons so the developer knows why generation was refused. This is a provider-side policy rejection, not an SDK malfunction.

Source

Thrown at packages/amazon-bedrock/src/amazon-bedrock-image-model.ts:279

      body: args,
      failedResponseHandler: createJsonErrorResponseHandler({
        errorSchema: AmazonBedrockErrorSchema,
        errorToMessage: error => `${error.type}: ${error.message}`,
      }),
      successfulResponseHandler: createJsonResponseHandler(
        amazonBedrockImageResponseSchema,
      ),
      abortSignal,
      fetch: this.config.fetch,
    });

    // Handle moderated/blocked requests
    if (response.status === 'Request Moderated') {
      const moderationReasons = response.details?.['Moderation Reasons'];
      const reasons = Array.isArray(moderationReasons)
        ? moderationReasons
        : ['Unknown'];
      throw new Error(
        `Amazon Bedrock request was moderated: ${reasons.join(', ')}`,
      );
    }

    // Check if images are present
    if (!response.images || response.images.length === 0) {
      throw new Error(
        'Amazon Bedrock returned no images. ' +
          (response.status ? `Status: ${response.status}` : ''),
      );
    }

    return {
      images: response.images,
      warnings,
      response: {
        timestamp: currentDate,
        modelId: this.modelId,

View on GitHub (pinned to 69428b1f8b)

Solutions

Reword the prompt to remove policy-sensitive content and retry
Remove or replace reference images that may trigger moderation
Read the moderation reasons in the error message and adjust accordingly
If you believe it's a false positive, test the same prompt directly in the Bedrock console and contact AWS if confirmed

Example fix

// before
await generateImage({ model, prompt: 'graphic violent battle scene' });
// after
await generateImage({ model, prompt: 'epic fantasy warriors, stylized illustration' });

Defensive patterns

Strategy: try-catch

Validate before calling

// pre-screen prompts client-side against your content policy before sending

Try / catch

try {
  await generateImage({ model, prompt });
} catch (e) {
  if (e instanceof Error && e.message.startsWith('Amazon Bedrock request was moderated')) {
    // surface the reasons to the user / sanitize the prompt
  }
}

Prevention

Screen prompts for policy-sensitive content before calling the API
Show moderation reasons from the error message to end users
Keep reference images policy-compliant

When it happens

Trigger: generateImage with a bedrock image model where the prompt (or reference images) trips Amazon Bedrock's content moderation guardrails, producing response.status === 'Request Moderated'.

Common situations: Prompts containing violence, sexual content, real-person likenesses, or other policy-sensitive content; occasionally false positives on borderline wording; regional policy differences.

Related errors

AI-assisted analysis of vercel/ai@69428b1f8b (2026-08-30). Data as JSON: /api/errors/df5d53c602137a81. Report an issue: GitHub.