AWS nova documentation change
Summary
Updated documentation for Responsible AI Toolkit with expanded content moderation configurations, availability details, and implementation guidance
Security assessment
Changes enhance documentation of security features (content moderation controls and responsible AI guidelines) but show no evidence of addressing a specific vulnerability. The updates clarify security-related configurations without indicating any security incident.
Diff
diff --git a/nova/latest/userguide/nova-responsible-ai-toolkit.md b/nova/latest/userguide/nova-responsible-ai-toolkit.md index 757d85362..c24141852 100644 --- a//nova/latest/userguide/nova-responsible-ai-toolkit.md +++ b//nova/latest/userguide/nova-responsible-ai-toolkit.md @@ -5 +5 @@ -Responsible AI toolkitCustomizable content moderation +Responsible AI ToolkitCustomizable content moderation @@ -7 +7 @@ Responsible AI toolkitCustomizable content moderation -# Responsible AI toolkit and content moderation +# Responsible AI Toolkit and content moderation @@ -9 +9 @@ Responsible AI toolkitCustomizable content moderation -## Responsible AI toolkit +## Responsible AI Toolkit @@ -11 +11 @@ Responsible AI toolkitCustomizable content moderation -Nova Forge provides a Responsible AI toolkit that includes training and evaluation data to align models to Amazon Nova's responsible AI guidelines during the training process, and runtime controls to moderate model responses during inference. +Nova Forge provides a Responsible AI Toolkit that includes training and evaluation data to align models to Amazon Nova's responsible AI guidelines during the training process, and runtime controls to moderate model responses during inference. @@ -25 +25,12 @@ Customizable content moderation settings (CCMS) allow adjustment of controls rel -These settings allow content generation through three available combinations: security; a combined setting for safety, sensitive content, and fairness; or all combinations together. +These settings allow content generation through three available configurations: + + * Security only + + * Safety, sensitive content, and fairness combined + + * All categories combined + + + + +The four content moderation categories are: @@ -38 +49,19 @@ These settings allow content generation through three available combinations: se -Regardless of CCMS configuration, Amazon Nova enforces essential, non-configurable controls to ensure responsible use of AI, such as controls to prevent harm to children and preserve privacy. +Regardless of your CCMS configuration, Amazon Nova enforces essential, non-configurable controls to ensure responsible use of AI, such as controls to prevent harm to children and preserve privacy. + +When using CCMS, we recommend using Continued Pre-Training (CPT) and starting from a pre-RAI alignment checkpoint (partially or fully pre-trained text-only) rather than the fully-aligned production checkpoint. These checkpoints have not undergone safety training or been steered toward specific RAI behaviors, allowing you to customize them more efficiently to your content moderation requirements. + +###### Tip + +When using CCMS with data mixing, consider adjusting the "rai" category percentage in your nova_data configuration to align with your specific content moderation requirements. + +### Availability + +CCMS is currently available for approved customers using: + + * Amazon Nova Lite 1.0 and Pro 1.0 models + + * Amazon Bedrock On-Demand inference + + * The us-east-1 (N. Virginia) region + + @@ -40 +68,0 @@ Regardless of CCMS configuration, Amazon Nova enforces essential, non-configurab -When using CCMS, use Continued Pre-Training (CPT) and start from a pre-RAI alignment checkpoint (partially or fully pre-trained text-only) rather than the fully-aligned production checkpoint. These checkpoints have not undergone safety training or been steered toward specific RAI behaviors, allowing more efficient customization to content moderation requirements. @@ -42 +70 @@ When using CCMS, use Continued Pre-Training (CPT) and start from a pre-RAI align -CCMS is currently available for Amazon Nova Lite 1.0 and Pro 1.0 with Amazon Bedrock On Demand inference in the us-east-1 (N. Virginia) region. To enable CCMS for Forge models, contact an Amazon Web Services account manager. +To enable CCMS for your Forge models, contact an Amazon Web Services account manager. @@ -50 +78 @@ To use the Amazon Web Services Documentation, Javascript must be enabled. Please -Reinforcement Learning +Supervised Fine-Tuning