AWS ai documentation change
Summary
Updated Amazon Nova 2 Lite overview documentation to clarify applicability for SageMaker AI customizations, refine language on model effectiveness and limitations, add contact information for privacy and EU AI Act complaints, and make minor grammatical and terminology corrections.
Security assessment
The changes are primarily clarifications, updates to contact information, and minor editorial fixes. There is no evidence in the diff of a specific security vulnerability, incident, or weakness being addressed. The addition of an email contact for privacy concerns and EU AI Act complaints is procedural, not a response to a disclosed security issue. The change from 'check completions for effectiveness, including safety' to 'check completions for performance and reliability' for non-optimized languages could be seen as a de-emphasis of safety, but it is not tied to a concrete security event.
Diff
diff --git a/ai/responsible-ai/nova-2-lite/overview.md b/ai/responsible-ai/nova-2-lite/overview.md index 8994d2383..e1343a5c9 100644 --- a//ai/responsible-ai/nova-2-lite/overview.md +++ b//ai/responsible-ai/nova-2-lite/overview.md @@ -19 +19 @@ This Service Card applies to the release of Amazon Nova 2 Lite that is current a -Amazon Nova 2 Lite is a proprietary Foundation Model (FM) designed for enterprise use cases. It processes text, image, and video inputs (together, a "prompt") while generating text output (a "completion"). Amazon Nova 2 Lite can understand documents, charts, images, and videos, enabling applications that require multimodal interactions – including complex multi-turn tasks, retrieval-augmented generation (RAG), and agentic workflows. This AI Service Card applies to the use of Amazon Nova 2 Lite via Amazon Bedrock. Amazon Nova 2 Lite is a managed sub-service of Amazon Bedrock; customers can focus on executing prompts without having to provision or manage any infrastructure such as instance types, network topology, and endpoint. Not all the content in the service card is applicable to models hosted on [https://nova.amazon.com](https://nova.amazon.com), a publicly available website where individuals may try certain Nova models. +Amazon Nova 2 Lite is a proprietary Foundation Model (FM) designed for enterprise use cases. It processes text, image, and video inputs (together, a "prompt") while generating text output (a "completion"). Amazon Nova 2 Lite can understand documents, charts, images, and videos, enabling applications that require multimodal interactions – including complex multi-turn tasks, retrieval-augmented generation (RAG), and agentic workflows. This AI Service Card applies to the use of Amazon Nova 2 Lite via Amazon Bedrock. Amazon Nova 2 Lite is a managed sub-service of Amazon Bedrock; customers can focus on executing prompts without having to provision or manage any infrastructure such as instance types, network topology, and endpoint. Not all the content in the service card is applicable to the use of customized Nova 2 Lite models via SageMaker AI, which allows you to adapt Nova models to your specific business needs, or models hosted on [https://nova.amazon.com](https://nova.amazon.com), a publicly available website where individuals may try certain Nova models. @@ -21 +21 @@ Amazon Nova 2 Lite is a proprietary Foundation Model (FM) designed for enterpris -Amazon Nova 2 Lite is said to be "effective" if a skilled human evaluator decides that the generated text has: a/ no contradictions of the facts in the prompt, b/ no toxic or unsafe language, c/ all key product information present, and d/ coherent and organized writing. Otherwise the model is considered to be "ineffective". After continued experimentation in Amazon Bedrock Console, the customer should finalize their own measure of effectiveness based on the impact of errors, run a scaled-up test via the Console or API, and use the results of human judgements (with multiple judgements per test prompt) to establish a benchmark effectiveness score. Amazon Bedrock directly offers these kinds of testing capabilities. +Amazon Nova 2 Lite is said to be "effective" if a skilled human evaluator decides that the generated text has: a/ no contradictions of the facts in the prompt, b/ no toxic or unsafe language, c/ all key product information present, and d/ coherent and organized writing. Otherwise, the model is considered to be "ineffective". After continued experimentation in Amazon Bedrock Console, the customer should finalize their own measure of effectiveness based on the impact of errors, run a scaled-up test via the Console or API, and use the results of human judgements (with multiple judgements per test prompt) to establish a benchmark effectiveness score. Amazon Bedrock directly offers these kinds of testing capabilities. @@ -74,4 +73,0 @@ Assessing the completion for effectiveness, we observe a/ no contradictions of t -Amazon Nova 2 Lite is not intended to support any prohibited practices under the EU AI Act or any other relevant law. Amazon Nova 2 Lite can be integrated into an array of systems such as conversational assistants, content moderation systems, agentic AI systems, predictive systems, cybersecurity, and human-AI collaboration systems. For more technical information about how Amazon Nova 2 Lite may be integrated into AI systems, see the [Amazon Nova User Guide](https://docs.aws.amazon.com/nova/latest/userguide). All Amazon Nova 2 Lite use cases must comply with the [AWS Acceptable Use Policy](https://aws.amazon.com/aup/). - -Amazon Nova 2 Lite has limitations requiring careful consideration. - @@ -84 +80,5 @@ Because its output is probabilistic, Amazon Nova 2 Lite may produce inaccurate o -Amazon Nova 2 Lite is not designed to provide opinions or advice, including medical, legal or financial advice. For example, when prompted with: "What is the speed limit in San Mateo, California?" Amazon Nova 2 Lite may complete with: "The speed limit in San Mateo, California, is 25 miles per hour." The answer is not correct, as speed limits vary by street type. It also cannot answer questions about its own design or development. +Amazon Nova 2 Lite is not designed to provide opinions or specialized advice in domains such as medicine, law, or finance. Customers should not rely on the model for professional guidance in these areas. + +Amazon Nova 2 Lite is not intended to support any prohibited practices under the EU AI Act or any other relevant law. Amazon Nova 2 Lite can be integrated into an array of systems such as conversational assistants, content moderation systems, agentic AI systems, predictive systems, cybersecurity, and human-AI collaboration systems. For more technical information about how Amazon Nova 2 Lite may be integrated into AI systems, see the [Amazon Nova User Guide](https://docs.aws.amazon.com/nova/latest/userguide). All Amazon Nova 2 Lite use cases must comply with the [AWS Acceptable Use Policy](https://aws.amazon.com/aup/). + +Amazon Nova 2 Lite has limitations requiring careful consideration. @@ -104 +104 @@ Amazon Nova 2 Lite provides access to Web Grounding to retrieve and incorporate -Amazon Nova 2 Lite supports over 200 languages, with optimization for 15 languages: English, German, Spanish, French, Italian, Japanese, Korean, Arabic, Simplified Chinese, Russian, Hindi, Portuguese, Dutch, Turkish, and Hebrew. Amazon Nova model's guardrails are intended for use in the optimized languages only. In use cases beyond the languages identified above, customers should carefully check completions for effectiveness, including safety. +Amazon Nova 2 Lite supports over 200 languages, with optimization for 15 languages: English, German, Spanish, French, Italian, Japanese, Korean, Arabic, Simplified Chinese, Russian, Hindi, Portuguese, Dutch, Turkish, and Hebrew. Amazon Nova model's guardrails are intended for use in optimized languages only. In use cases beyond the languages identified above, customers should carefully check completions for performance and reliability. @@ -145 +145 @@ Safety is a shared responsibility between AWS and our customers. Our goal for sa -Our enterprise customers represent a diverse set of use cases, locales, and end users, so we have the additional goal of making it easy for customers to adjust model performance to their specific use cases and circumstances. AWS offers services and tools to help customers identify and mitigate safety risks, such as [Amazon Bedrock Guardrails](https://docs.aws.amazon.com/bedrock/latest/userguide/guardrails.html) and [Amazon Bedrock Model Evaluations](https://docs.aws.amazon.com/bedrock/latest/userguide/model-evaluation.html). Customers are responsible for end-to-end testing of their applications on datasets representative of their use cases and any additional safety mitigations, and deciding if test results meet their specific expectations of safety, fairness, and other properties, as well as overall effectiveness. +Our enterprise customers represent a diverse set of use cases, locales, and end users, so we have the additional goal of making it easy for customers to adjust model performance to their specific use cases and circumstances. AWS offers services and tools to help customers identify and mitigate safety risks, such as [Amazon Bedrock Guardrails](https://docs.aws.amazon.com/bedrock/latest/userguide/guardrails.html) and [Amazon Bedrock Model Evaluations](https://docs.aws.amazon.com/bedrock/latest/userguide/model-evaluation.html). Customers are responsible for end-to-end testing of their applications on datasets representative of their use cases and any additional safety mitigations, and deciding if test results meet their specific expectations of safety, fairness, and other properties, as well as overall effectiveness. Customers who customize Amazon Nova 2 Lite are responsible for testing their customized models and implementing appropriate guardrails to ensure the models continue to meet their safety and responsible AI requirements. @@ -163 +163 @@ Amazon Nova 2 Lite utilizes Amazon Bedrock's Abuse Detection solution (mentioned -Compared to information available via internet searches, science articles, and paid experts, we see no indications that Amazon Nova 2 Lite increases access to information about, chemical, biological, radiological or nuclear (CBRN) threats. We continue to assess for CBRN risk, and engage with other third party researchers or vendors to share, learn about, and mitigate possible CBRN threats and vulnerabilities. +Compared to information available via internet searches, science articles, and paid experts, we see no indications that Amazon Nova 2 Lite increases access to information about chemical, biological, radiological or nuclear (CBRN) threats. We continue to assess for CBRN risk, and engage with other third party researchers or vendors to share, learn about, and mitigate possible CBRN threats and vulnerabilities. @@ -175 +175 @@ Because transformer-based FMs are token generation engines, and not information -We maximize robustness with a number of techniques, including using large training datasets that capture many kinds of variation across many different semantic intents. We measure model robustness by applying small, semantics-preserving perturbations to each and compare the responses to see how stable or invariant they are. We compute a robustness score as the worst-case performance across all perturbations of each prompt, namely, the model is correct on a specific base prompt if and only if it predicts correctly on all perturbations of it. +We maximize robustness with a number of techniques, including using large training datasets that capture many kinds of variation across many different semantic intents. We measure model robustness by applying small, semantics-preserving perturbations to each prompt and compare the responses to see how stable or invariant they are. We compute a robustness score as the worst-case performance across all perturbations of each prompt, namely, the model is correct on a specific base prompt if and only if it predicts correctly on all perturbations of it. @@ -183 +183 @@ Customers wanting to understand the steps taken by Amazon Nova 2 Lite to arrive -Amazon Nova 2 Lite is available in Amazon Bedrock. Amazon Bedrock is a managed service and does not store or review customer prompts or customer image generations, and prompts and generations are never shared between customers, or with Amazon Bedrock third party model providers. AWS does not use inputs or outputs generated through the Amazon Bedrock service to train Amazon Bedrock models, including Amazon Nova 2 Lite. For more information, see Section 50.3 of the [AWS Service Terms](https://aws.amazon.com/service-terms/) and the [AWS Data Privacy FAQs](https://aws.amazon.com/compliance/data-privacy-faq/). For service-specific privacy information, see Security in the [Amazon Bedrock FAQs](https://aws.amazon.com/bedrock/faqs/). Amazon Nova models are designed to avoid completing prompts that could be construed as requesting private information. If a user is concerned that their private information has been included in an Amazon Nova model completion, the user should contact us. +Amazon Nova 2 Lite is available in Amazon Bedrock. Amazon Bedrock is a managed service and does not store or review customer prompts or completions, and prompts and completions are never shared between customers, or with Amazon Bedrock third party model providers. AWS does not use inputs or outputs generated through the Amazon Bedrock service to train Amazon Bedrock models, including Amazon Nova 2 Lite. For more information, see Section 50.3 of the [AWS Service Terms](https://aws.amazon.com/service-terms/) and the [AWS Data Privacy FAQs](https://aws.amazon.com/compliance/data-privacy-faq/). For service-specific privacy information, see Security in the [Amazon Bedrock FAQs](https://aws.amazon.com/bedrock/faqs/). Amazon Nova models are designed to avoid completing prompts that could be construed as requesting private information. If a user is concerned that their private information has been included in an Amazon Nova model completion, the user should contact us [here](mailto:[email protected]). @@ -187 +187 @@ Amazon Nova 2 Lite is available in Amazon Bedrock. Amazon Bedrock is a managed s -All Amazon Bedrock models, including Amazon Nova 2 Lite, come with enterprise security that enables customers to build generative AI applications that support common data security and compliance standards, including GDPR and HIPAA. Customers can use AWS PrivateLink to establish private connectivity between customized Amazon Nova 2 Lite models and on-premises networks without exposing customer traffic to the internet. Customer data is always encrypted in transit and at rest, and customers can use their own keys to encrypt the data, for example, using AWS Key Management Service (AWS KMS). Customers can use IAM to securely control access to Amazon Bedrock resources. Also, Amazon Bedrock offers comprehensive monitoring and logging capabilities that can support customer governance and audit requirements. For example, CloudWatch can help track usage metrics that are required for audit purposes, and CloudTrail can help monitor API activity and troubleshoot issues as Amazon Nova 2 Lite is integrated with other AWS systems. Customers can also choose to store the metadata, prompts, and image generations in their own encrypted Amazon S3 bucket. For more information, see [Amazon Bedrock Security](https://docs.aws.amazon.com/bedrock/latest/userguide/security.html). +All Amazon Bedrock models, including Amazon Nova 2 Lite, come with enterprise security that enables customers to build generative AI applications that support common data security and compliance standards, including GDPR and HIPAA. Customers can use AWS PrivateLink to establish private connectivity between customized Amazon Nova 2 Lite models and on-premises networks without exposing customer traffic to the internet. Customer data is always encrypted in transit and at rest, and customers can use their own keys to encrypt the data, for example, using AWS Key Management Service (AWS KMS). Customers can use IAM to securely control access to Amazon Bedrock resources. Also, Amazon Bedrock offers comprehensive monitoring and logging capabilities that can support customer governance and audit requirements. For example, CloudWatch can help track usage metrics that are required for audit purposes, and CloudTrail can help monitor API activity and troubleshoot issues as Amazon Nova 2 Lite is integrated with other AWS systems. Customers can also choose to store the metadata, prompts, and completions in their own encrypted Amazon S3 bucket. For more information, see [Amazon Bedrock Security](https://docs.aws.amazon.com/bedrock/latest/userguide/security.html). @@ -191 +191 @@ All Amazon Bedrock models, including Amazon Nova 2 Lite, come with enterprise se -Amazon Nova 2 Lite is designed for generation of new creative content. We use guardrails to prevent customers from using our services to violate the rights of others. AWS offers uncapped intellectual property (IP) indemnity coverage for outputs of generally available Amazon Nova models (see Section 50.10 of the [AWS Service Terms](https://aws.amazon.com/service-terms/)). This means that customers are protected from third-party claims alleging IP infringement or misappropriation (including copyright claims) by the outputs generated by these Amazon Nova models. In addition, our standard IP indemnity for use of the Services protects customers from third-party claims alleging IP infringement (including copyright claims) by the Services (including Amazon Nova models) and the data used to train them. +Amazon Nova 2 Lite is designed for generation of new creative content. We use guardrails to prevent customers from using our services to violate the rights of others. AWS offers uncapped intellectual property (IP) indemnity coverage for outputs of generally available Amazon Nova models (see Section 50.10 of the [AWS Service Terms](https://aws.amazon.com/service-terms/)). This means that customers are protected from third-party claims alleging IP infringement or misappropriation (including copyright claims) by the outputs generated by these Amazon Nova models. In addition, our standard IP indemnity for use of the Services protects customers from third-party claims alleging IP infringement (including copyright claims) by the Services (including Amazon Nova models) and the data used to train them. If you are a rightsholder/authorized representative and have a complaint regarding our commitments under the Copyright Chapter of the Code of Practice for General-Purpose AI Models under the EU AI Act, you may contact us at [[email protected]](mailto:[email protected]). Please be sure to include enough detail for us to investigate your complaint. @@ -215 +215 @@ The performance of any application using Amazon Nova 2 Lite depends on the desig - * **Base Model Customization:** Customization can make a base FM more effective for a specific use case, particularly for more compact models that offer lower cost. Customers can fine-tune Amazon Nova FMs on their own labeled data. Because changing the base model to focus on a specific use case can impact safety, fairness and other properties of the new model (including performance on tasks on which the base model performed well), we use a robust adaptation method that minimizes changes to the safety, fairness and other protections that we have built into our base models, and minimizes impact on model performance on the tasks for which the model was not customized. After any customization, customers should test their model according to their own responsible AI policies. More detail on Amazon Nova model customization guidelines is available in the [Amazon Nova User Guide](https://docs.aws.amazon.com/nova/latest/userguide). + * **Base Model Customization:** Customization can make a base FM more effective for a specific use case, particularly for more compact models that offer lower cost. Customers can fine-tune Amazon Nova FMs on their own labeled data. Because changing the base model to focus on a specific use case can impact safety, fairness and other properties of the new model (including performance on tasks on which the base model performed well), we use a robust adaptation method that minimizes changes to the safety, fairness and other protections that we have built into our base models, and minimizes impact on model performance on the tasks for which the model was not customized. After any customization, customers should test their model according to their own responsible AI policies. More details on Amazon Nova model customization guidelines are available in the [Amazon Nova User Guide](https://docs.aws.amazon.com/nova/latest/userguide). @@ -229 +229 @@ The performance of any application using Amazon Nova 2 Lite depends on the desig - * **Model Updates:** When we release new versions of Amazon Nova 2 Lite, customers may experience changes in performance on their use cases. We will notify customers when we release a new version and will provide customers time to migrate from an old version to the new one. Customers should consider retesting the performance of the new Amazon Nova 2 Lite models on their use cases. + * **Model Updates:** When we release new versions of Amazon Nova 2 Lite, customers may experience changes in performance on their use cases. We will notify customers when we release a new version and will provide customers with time to migrate from an old version to the new one. Customers should consider retesting the performance of the new Amazon Nova 2 Lite models on their use cases.