AWS general documentation change
Summary
Added new guardrail quotas for content policies, contextual grounding, sensitive information filters, topic policies, and word filters. Updated token quotas for Anthropic Claude models and bedrock-mantle endpoints.
Security assessment
The changes document security-related guardrail features (content filtering, sensitive information detection, contextual grounding) but show no evidence of addressing a specific vulnerability. The updates describe operational quotas and recommendations for security features without indicating any security incident resolution.
Diff
diff --git a/general/latest/gr/bedrock.md b/general/latest/gr/bedrock.md index 98ee233b5..d17899a10 100644 --- a//general/latest/gr/bedrock.md +++ b//general/latest/gr/bedrock.md @@ -285,0 +286,3 @@ Name | Default | Adjustable | Description +(Guardrails) Content policy maximum input size in text units (Classic tier) | Each supported Region: 106 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-B90A0AF2) | The maximum size of input text, measured in text units, that can be processed for content filters. While this limit applies to the classic tier, we recommend migrating to standard tier due to its superior robustness, additional capabilities, and multi-lingual support. +(Guardrails) Content policy maximum input size in text units (Standard tier - Recommended) | Each supported Region: 106 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-C0BA182F) | The maximum size of input text, measured in text units, that can be processed for content filters. This applies to the standard tier, which is recommended. +(Guardrails) Contextual grounding policy maximum input size in text units | Each supported Region: 106 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-392F9C48) | The maximum size of input text, measured in text units, that can be processed by Contextual grounding policies @@ -290,0 +294,2 @@ Name | Default | Adjustable | Description +(Guardrails) On-demand ApplyGuardrail Content filter policy text units burst rate (Classic tier) | Each supported Region: 25 | No | The maximum number of text units in one burst that can be processed for content filters. While this limit applies to the classic tier, we recommend migrating to standard tier due to its superior robustness, additional capabilities, and multi-lingual support. +(Guardrails) On-demand ApplyGuardrail Content filter policy text units burst rate (Standard tier - Recommended) | Each supported Region: 25 | No | The maximum number of text units in one burst that can be processed for content filters. This applies to the standard tier, which is recommended. @@ -292,0 +298,2 @@ Name | Default | Adjustable | Description +(Guardrails) On-demand ApplyGuardrail Denied topic policy text units burst rate (Classic tier) | Each supported Region: 25 | No | The maximum number of text units in one burst that can be processed for denied topics. While this limit applies to the classic tier, we recommend migrating to standard tier due to its superior robustness, additional capabilities, and multi-lingual support. +(Guardrails) On-demand ApplyGuardrail Denied topic policy text units burst rate (Standard tier - Recommended) | Each supported Region: 25 | No | The maximum number of text units in one burst that can be processed for denied topics. This applies to the standard tier, which is recommended. @@ -294,0 +302 @@ Name | Default | Adjustable | Description +(Guardrails) On-demand ApplyGuardrail Sensitive information filter policy text units burst rate | Each supported Region: 25 | No | The maximum number of text units in one burst that can be processed for sensitive information filters. @@ -295,0 +304 @@ Name | Default | Adjustable | Description +(Guardrails) On-demand ApplyGuardrail Word filter policy text units burst rate | Each supported Region: 25 | No | The maximum number of text units in one burst that can be processed for word filters. @@ -296,0 +306 @@ Name | Default | Adjustable | Description +(Guardrails) On-demand ApplyGuardrail contextual grounding policy text units burst rate | Each supported Region: 106 | No | The maximum number of text units in one burst that can be processed for contextual grounding. @@ -297,0 +308 @@ Name | Default | Adjustable | Description +(Guardrails) On-demand ApplyGuardrail requests burst rate | Each supported Region: 25 | No | The maximum number of ApplyGuardrail API calls that you can send in one burst. @@ -301,0 +313,3 @@ Name | Default | Adjustable | Description +(Guardrails) Sensitive information policy maximum input size in text units | Each supported Region: 106 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-437802FD) | The maximum size of input text, measured in text units, that can be processed by Sensitive information filter policies +(Guardrails) Topic policy maximum input size in text units (Classic tier) | Each supported Region: 106 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-DD0273B7) | The maximum size of input text, measured in text units, that can be processed for denied topics. While this limit applies to the classic tier, we recommend migrating to standard tier due to its superior robustness, additional capabilities, and multi-lingual support. +(Guardrails) Topic policy maximum input size in text units (Standard tier - Recommended) | Each supported Region: 106 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-3B60C1ED) | The maximum size of input text, measured in text units, that can be processed for denied topics. This applies to the standard tier, which is recommended. @@ -304,0 +319 @@ Name | Default | Adjustable | Description +(Guardrails) Word policy maximum input size in text units | Each supported Region: 106 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-A88E7250) | The maximum size of input text, measured in text units, that can be processed by Word filter policies @@ -646 +661 @@ Cross-region model inference tokens per minute for Anthropic Claude Opus 4.6 V1 -Cross-region model inference tokens per minute for Anthropic Claude Opus 4.7 | ca-central-1: 10,000,000 eu-central-1: 10,000,000 Each of the other supported Regions: 30,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-5DB28B7B) | The maximum number of cross-region tokens that you can submit for model inference in one minute for Anthropic Claude Opus 4.7. The quota considers the combined sum of input and output tokens across all requests to Converse, ConverseStream, InvokeModel and InvokeModelWithResponseStream. +Cross-region model inference tokens per minute for Anthropic Claude Opus 4.7 | Each supported Region: 10,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-5DB28B7B) | The maximum number of cross-region tokens that you can submit for model inference in one minute for Anthropic Claude Opus 4.7. The quota considers the combined sum of input and output tokens across all requests to Converse, ConverseStream, InvokeModel and InvokeModelWithResponseStream. @@ -652,0 +668 @@ Cross-region model inference tokens per minute for Anthropic Claude Sonnet 4.6 | +Cross-region model inference tokens per minute for Anthropic Claude Sonnet 5 | Each supported Region: 6,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-D4FBCF4E) | The maximum number of cross-region tokens that you can submit for model inference in one minute for Anthropic Claude Sonnet 5. The quota considers the combined sum of input and output tokens across all requests to Converse, ConverseStream, InvokeModel and InvokeModelWithResponseStream. @@ -697 +713 @@ Global cross-region model inference tokens per day for Anthropic Claude Opus 4.6 -Global cross-region model inference tokens per day for Anthropic Claude Opus 4.7 | ca-central-1: 14,400,000,000 eu-central-1: 14,400,000,000 Each of the other supported Regions: 43,200,000,000 | No | The maximum number of global cross-region tokens that you can submit for model inference in one day for Anthropic Claude Opus 4.7. The quota considers the combined sum of input and output tokens across all requests to Converse, ConverseStream, InvokeModel and InvokeModelWithResponseStream. +Global cross-region model inference tokens per day for Anthropic Claude Opus 4.7 | Each supported Region: 14,400,000,000 | No | The maximum number of global cross-region tokens that you can submit for model inference in one day for Anthropic Claude Opus 4.7. The quota considers the combined sum of input and output tokens across all requests to Converse, ConverseStream, InvokeModel and InvokeModelWithResponseStream. @@ -702,0 +719 @@ Global cross-region model inference tokens per day for Anthropic Claude Sonnet 4 +Global cross-region model inference tokens per day for Anthropic Claude Sonnet 5 | Each supported Region: 8,640,000,000 | No | The maximum number of global cross-region tokens that you can submit for model inference in one day for Anthropic Claude Sonnet 5. The quota considers the combined sum of input and output tokens across all requests to Converse, ConverseStream, InvokeModel and InvokeModelWithResponseStream. @@ -711 +728 @@ Global cross-region model inference tokens per minute for Anthropic Claude Opus -Global cross-region model inference tokens per minute for Anthropic Claude Opus 4.7 | ca-central-1: 10,000,000 eu-central-1: 10,000,000 Each of the other supported Regions: 30,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-34152C1D) | The maximum number of global cross-region tokens that you can submit for model inference in one minute for Anthropic Claude Opus 4.7. The quota considers the combined sum of input and output tokens across all requests to Converse, ConverseStream, InvokeModel and InvokeModelWithResponseStream. +Global cross-region model inference tokens per minute for Anthropic Claude Opus 4.7 | Each supported Region: 10,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-34152C1D) | The maximum number of global cross-region tokens that you can submit for model inference in one minute for Anthropic Claude Opus 4.7. The quota considers the combined sum of input and output tokens across all requests to Converse, ConverseStream, InvokeModel and InvokeModelWithResponseStream. @@ -716,0 +734 @@ Global cross-region model inference tokens per minute for Anthropic Claude Sonne +Global cross-region model inference tokens per minute for Anthropic Claude Sonnet 5 | Each supported Region: 6,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-DD84E5CA) | The maximum number of global cross-region tokens that you can submit for model inference in one minute for Anthropic Claude Sonnet 5. The quota considers the combined sum of input and output tokens across all requests to Converse, ConverseStream, InvokeModel and InvokeModelWithResponseStream. @@ -816 +834 @@ Model invocation max tokens per day for Anthropic Claude Opus 4.6 V1 (doubled fo -Model invocation max tokens per day for Anthropic Claude Opus 4.7 (doubled for cross-region calls) | ca-central-1: 7,200,000,000 eu-central-1: 7,200,000,000 Each of the other supported Regions: 21,600,000,000 | No | Daily maximum tokens for model inference for Anthropic Claude Opus 4.7. Combines sum of input and output tokens across all requests to Converse, ConverseStream, InvokeModel and InvokeModelWithResponseStream. Doubled for cross-region calls; not applicable in case of approved TPM increase. +Model invocation max tokens per day for Anthropic Claude Opus 4.7 (doubled for cross-region calls) | Each supported Region: 7,200,000,000 | No | Daily maximum tokens for model inference for Anthropic Claude Opus 4.7. Combines sum of input and output tokens across all requests to Converse, ConverseStream, InvokeModel and InvokeModelWithResponseStream. Doubled for cross-region calls; not applicable in case of approved TPM increase. @@ -822,0 +841 @@ Model invocation max tokens per day for Anthropic Claude Sonnet 4.6 (doubled for +Model invocation max tokens per day for Anthropic Claude Sonnet 5 (doubled for cross-region calls) | Each supported Region: 4,320,000,000 | No | Daily maximum tokens for model inference for Anthropic Claude Sonnet 5. Combines sum of input and output tokens across all requests to Converse, ConverseStream, InvokeModel and InvokeModelWithResponseStream. Doubled for cross-region calls; not applicable in case of approved TPM increase. @@ -1392,4 +1411,4 @@ UpdateAgentKnowledgeBase requests per second | Each supported Region: 4 | No | T -[bedrock-mantle endpoint] Input tokens per minute for GPT-5.5 | Each supported Region: 5,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-2555C05B) | [bedrock-mantle endpoint] The maximum number of input tokens per minute your account can submit to GPT-5.5 in this region. Shared across all APIs for this model. -[bedrock-mantle endpoint] Output tokens per minute for Claude Opus 4.7 | Each supported Region: 4,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-15CD18AC) | [bedrock-mantle endpoint] The maximum number of output tokens per minute Claude Opus 4.7 can generate for your account in this region. Shared across all APIs for this model. -[bedrock-mantle endpoint] Output tokens per minute for Claude Opus 4.8 | Each supported Region: 4,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-37491D63) | [bedrock-mantle endpoint] The maximum number of output tokens per minute Claude Opus 4.8 can generate for your account in this region. Shared across all APIs for this model. -[bedrock-mantle endpoint] Output tokens per minute for GPT-5.4 | Each supported Region: 4,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-08036ECB) | [bedrock-mantle endpoint] The maximum number of output tokens per minute GPT-5.4 can generate for your account in this region. Shared across all APIs for this model. +[bedrock-mantle endpoint] Input tokens per minute for GPT-5.5 | Each supported Region: 10,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-2555C05B) | [bedrock-mantle endpoint] The maximum number of input tokens per minute your account can submit to GPT-5.5 in this region. Shared across all APIs for this model. +[bedrock-mantle endpoint] Output tokens per minute for Claude Opus 4.7 | Each supported Region: 2,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-15CD18AC) | [bedrock-mantle endpoint] The maximum number of output tokens per minute Claude Opus 4.7 can generate for your account in this region. Shared across all APIs for this model. +[bedrock-mantle endpoint] Output tokens per minute for Claude Opus 4.8 | Each supported Region: 2,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-37491D63) | [bedrock-mantle endpoint] The maximum number of output tokens per minute Claude Opus 4.8 can generate for your account in this region. Shared across all APIs for this model. +[bedrock-mantle endpoint] Output tokens per minute for GPT-5.4 | Each supported Region: 2,000,000 | [Yes](https://console.aws.amazon.com/servicequotas/home/services/bedrock/quotas/L-08036ECB) | [bedrock-mantle endpoint] The maximum number of output tokens per minute GPT-5.4 can generate for your account in this region. Shared across all APIs for this model.