Amazon Bedrock FAQs changed AWS, May 12, 2025

1. +Added: Which FMs are available in Amazon Bedrock?
2. +Added: Amazon Bedrock customers can choose from some of the most cutting-edge FMs available today. This includes models from:
3. +Added: See here for supported foundation models from each provider:
4. -Removed: Which FMs are available on Amazon Bedrock?
5. -Removed: Amazon Bedrock customers can choose from some of the most cutting-edge FMs available today. Currently we offer 47 models. This includes language and embeddings models from:
6. -Removed: AI21: Jamba 1.5 Large, Jamba 1.5 Mini, Jamba-Instruct, Jurassic-2 Mid, Jurassic-2 Ultra
7. -Removed: Anthropic: Claude 3.5 Sonnet, Claude 3.5 Haiku, Claude 3 Opus, Claude 3 Haiku, Claude 3 Sonnet, Claude 2.1, Claude 2.0, Claude Instant
8. -Removed: Cohere: Command R+, Command R, Command, Command Light, Embed - English, Embed - Multilingual
9. -Removed: Meta: Llama 3.2 90B, Llama 3.2 11B, Llama 3.2 3B, Llama 3.2 1B, Llama 3.1 8B, Llama 3.1 70B, Llama 3.1 405B, Llama 3 8B, Llama 3 70B, Llama 2 13B, Llama 2 70B
10. -Removed: Mistral AI: Mistral Large 2 (24.07), Mistral Large (24.02), Mistral Small (24.02), Mixtral 8x7B, Mistral 7B
11. -Removed: Stability AI: Stable Image Ultra, Stable Diffusion 3 Large, Stable Image Core, Stable Diffusion XL 1.0
12. -Removed: Amazon: Amazon Titan Text Premier, Amazon Titan Text Express, Amazon Titan Text Lite, Amazon Titan Text Embeddings, Amazon Titan Text Embeddings V2, Amazon Titan Multimodal Embeddings, Amazon Titan Image Generator, Amazon Titan Image Generator v2
13. +Added: Amazon Bedrock Guardrails helps you implement safeguards for your generative AI applications based on your use cases and responsible AI policies. You can create multiple guardrails tailored to different use cases and apply them across multiple foundation models (FMs), providing a consistent user experience and standardizing safety and privacy controls across your generative AI applications.
14. -Removed: Amazon Bedrock Guardrails help you implement safeguards for your generative AI applications based on your use cases and responsible AI policies. Guardrails helps control the interaction between users and FMs by filtering undesirable and harmful content and will soon redact personally identifiable information (PII), enhancing content safety and privacy in generative AI applications. You can create multiple guardrails with different configurations tailored to specific use cases. Additionally, with the guardrails you can continually monitor and analyze user inputs and FM responses that might violate customer-defined policies.
15. +Added: Guardrails help you to define a set of six policies to help safeguard your generative AI applications. You can configure the following policies in Amazon Bedrock Guardrails:
16. +Added: Multi modal content filters - Configure thresholds to detect and filter harmful text and/or image content across categories including hate, insults, sexual, violence, misconduct, and prompt attacks.
17. +Added: Denied topics - Define a set of topics that are undesirable in the context of your application. The filter will help block them if detected in user queries or model responses.
18. +Added: Word filters - Configure filters to help block undesirable words, phrases, and profanity (exact match). Such words can include offensive terms, competitor names, etc.
19. +Added: Sensitive information filters - Configure filters to help block or mask sensitive information, such as personally identifiable information (PII), or custom regex in user inputs and model responses. Blocking or masking is done based on probabilistic detection of sensitive information in standard formats in entities such as SSN number, Date of Birth, address, etc. This also allows configuring regular expression based detection of patterns for identifiers.
20. +Added: Contextual grounding check - Help detect and filter hallucinations if the responses are not grounded (e.g., factually inaccurate or new information) in the source information and irrelevant to user's query or instruction.
21. +Added: Automated Reasoning checks - Help detect factual inaccuracies in generated content, suggest corrections, and explain why responses are accurate by checking against a structured, mathematical representation of knowledge called an Automated Reasoning Policy.
22. +Added: What modalities are supported with Bedrock Guardrails?
23. +Added: Bedrock Guardrails supports both text and image content to enable customers build secure generative AI applications at scale.
24. -Removed: Guardrails help you to define a set of policies to help safeguard your generative AI applications. You can configure the following policies in a guardrail.
25. -Removed: Contextual grounding checks: help detect and filter hallucinations if the responses are not grounded (e.g., factually inaccurate or new information) in the source information and irrelevant to user's query or instruction.
26. -Removed: Automated Reasoning checks: help detect factual inaccuracies in generated content, suggest corrections, and explain why responses are accurate by checking against a structured, mathematical representation of knowledge called an Automated Reasoning Policy.
27. -Removed: Content filters: help you configure thresholds to detect and filter harmful text content across categories such as hate, insults, sexual, violence, misconduct, and prompt attacks. Additionally, content filters can detect and filter harmful image content across these categories thereby helping build safe multimodal applications.
28. -Removed: Denied topics: help you define a set of topics that are undesirable in the context of your application. For example, an online banking assistant can be designed to refrain from providing investment advice.
29. -Removed: Word filters: help you define a set of words to block in user inputs and FM-generated responses.
30. -Removed: Sensitive information filter: helps you react sensitive information like a set of PII that can be redacted in FM-generated responses. Based on the use case, Guardrails can also help you block a user input if it contains PII.
31. +Added: Amazon Bedrock Guardrails works with a wide range of models including FMs supported in Amazon Bedrock, fine-tuned models, as well as self-hosted models outside Amazon Bedrock. User inputs and model outputs can be evaluated independently for third-party and self-hosted models using the ApplyGuardrail API. Amazon Bedrock Guardrails can also be integrated with Amazon Bedrock Agents and Amazon Bedrock Knowledge Bases to build safe and secure generative AI applications aligned with responsible AI policies
32. +Added: What languages are supported with Bedrock Guardrails?
33. +Added: Currently, Amazon Bedrock Guardrails supports English, French, and Spanish in natural language. Using any other language will result in ineffective results.
34. -Removed: Amazon Bedrock Guardrails works with a wide range of models including FMs supported in Amazon Bedrock, fine-tuned models, as well as, self-hosted models outside Amazon Bedrock. User inputs and model outputs can be evaluated independently for third-party and self-hosted models using the ApplyGuardrail API. Amazon Bedrock Guardrails can also be integrated with Amazon Bedrock Agents and Amazon Bedrock Knowledge Bases to build safe and secure generative AI applications aligned with responsible AI policies
35. -Removed: Does AWS offer an intellectual property indemnity covering copyright claims for its generative AI services?
36. -Removed: AWS offers an uncapped intellectual property (IP) indemnity for copyright claims arising from generative output of the following generally available Amazon generative AI services: Amazon Titan models, and other services listed in Section 50.10 of the Service Terms (the "Indemnified Generative AI Services"). This means that customers are protected from third-party claims alleging copyright infringement by the output generated by the Indemnified Generative AI Services in response to inputs or other data provided by the customer. Customers must also use the services responsibly, such as not inputting infringing data or disabling a service's filtering features.
37. +Added: How can I enforce Guardrails across my organization?
38. +Added: Amazon Bedrock Guardrails provides the ability to establish mandatory guardrails for every inference call using IAM policy-based enforcement capabilities. See here for details.
39. +Added: Does AWS offer an intellectual property indemnity covering copyright claims for its generative AI services?
40. +Added: AWS offers an uncapped intellectual property (IP) indemnity for copyright claims arising from generative output of the following generally available Amazon generative AI services: Amazon models, and other services listed in Section 50.10 of the Service Terms (the "Indemnified Generative AI Services"). This means that customers are protected from third-party claims alleging copyright infringement by the output generated by the Indemnified Generative AI Services in response to inputs or other data provided by the customer. Customers must also use the services responsibly, such as not inputting infringing data or disabling a service's filtering features.
41. +Added: Amazon Bedrock Guardrails offers Sensitive Information filters which provide 31 PIIs that include social security numbers and phone numbers. See here for the complete list.
42. +Added: What is the pricing model for using Amazon Bedrock Guardrails?
43. +Added: Amazon Bedrock Guardrails is priced on a per-use model for both text and image content. Please see the Guardrails pricing page for pricing details.
44. -Removed: Foundations model have native safeguards and they are the default protections associated with each model. These native safeguards are NOT part of Amazon Bedrock Guardrails. Amazon Bedrock Guardrails is an added layer of customized safeguards that can be optionally applied by the customer based on their application requirements and responsible AI policies.
45. -Removed: As part of Amazon Bedrock Guardrails, SSN and phone number detection are part of the 30+ off the shelf PIIs. Full list here .
46. -Removed: Is there a separate cost for customers to build custom Amazon Bedrock Guardrails? And it is applied to both the input and output?
47. -Removed: There is a separate cost for using Amazon Bedrock Guardrails. It can be applied for both input and output. Pricing at the bottom of the page here . The pricing for image support with content filters (currently in public preview) will be announced during general availability (GA).
48. +Added: What image formats are supported for multimodal content?
49. +Added: PNG and JPEG image formats are supported with Bedrock Guardrails.
50. +Added: Simply navigate to the Amazon Bedrock Model Catalog page in the Bedrock console where you can search for Amazon Bedrock Marketplace model listings along with the serverless Amazon Bedrock models. After you have selected the Amazon Bedrock Marketplace model you want to use, you can subscribe to the model through the Model Detail page, accepting the EULA and price(s) set by the provider. Once the subscription is complete, which typically takes a few minutes, you can deploy the model to a fully managed SageMaker endpoint by clicking on Deploy in the Model Detail page or by using APIs. In the deployment step, you can select your desired number of instances and instance types to meet your workload. Once the endpoint is setup, which typically takes 10 - 15 minutes, you can start making inference calls to the endpoint and use the model in Bedrock's advanced tools, provided the model is compatible with Bedrock's Converse API.
51. -Removed: Simply navigate to the Amazon Bedrock Model Catalgo page in the Bedrock console where you can search for Amazon Bedrock Marketplace model listings along with the serverless Amazon Bedrock models. After you have selected the Amazon Bedrock Marketplace model you want to use, you can subscribe to the model through the Model Detail page, accepting the EULA and price(s) set by the provider. Once the subscription is complete, which typically takes a few minutes, you can deploy the model to a fully managed SageMaker endpoint by clicking on Deploy in the Model Detail page or by using APIs. In the deployment step, you can select your desired number of instances and instance types to meet your workload. Once the endpoint is setup, which typically takes 10 - 15 minutes, you can start making inference calls to the endpoint and use the model in Bedrock's advanced tools, provided the model is compatible with Bedrock's Converse API.
52. +Added: Bedrock Data Automation supports PDF, PNG, JPG, TIFF, a max of 1500 pages, and a max file size of 500MB per API request. By default, BDA will support 50 concurrent jobs and 10 transactions per second per customer.
53. -Removed: Bedrock Data Automation will support PDF, PNG, JPG, TIFF, a max of 100 pages, and a max file size of 500MB per API request. BDA will support a max concurrency of 5 document packages and throughput of 1 page per second per customer.
54. +Added: Standard output will provide summarization, detected explicit content, detected text, logo detection and Ad taxonomy: IAB for images. Standard output will be used for BDA integration with Bedrock Knowledge Bases.
55. +Added: Bedrock Data Automation supports JPG, PNG, a max resolution of 4K, and a max file size of 5 MB per API request. By default, BDA supports a max concurrency of 20 images at 10 transactions per second (TPS) per customer.
56. -Removed: Standard output will provide summarization, detected explicit content, detected text, and Ad taxonomy: IAB for images. Standard output will be used for BDA integration with Bedrock Knowledge Bases.
57. -Removed: Bedrock Data Automation will support JPG, PNG, a max resolution of 4K, and a max file size of 5 MB per API request. BDA will support a max concurrency of 100 images at 1 image per second per customer.
58. +Added: Standard output will provide full video summary, chapter segmentation, chapter summary, full audio transcription, speaker identification, detected explicit content, detected text, logo detection and Interactive Advertising Bureau (IAB) taxonomy for videos. Full video summary is optimized for content with descriptive dialogue such as product overviews, trainings, news casts, and documentaries.
59. +Added: Bedrock Data Automation supports MOV and MKV with H.264, VP8, VP9, a max video duration of 4 hours, and a max file size of 2 GB per API request. By default, BDA supports a max concurrency of 20 videos at 10 transactions per second (TPS) per customer.
60. -Removed: Standard output will provide full video summary, scene segmentation, scene summary, full audio transcription, speaker identification, detected explicit content, detected text, and Interactive Advertising Bureau (IAB) taxonomy for videos. Full video summary is optimized for content with descriptive dialogue such as product overviews, trainings, news casts, and documentaries.

76 more changed paragraphs are on the page itself.