This page is Amazon's description of the data that Amazon.com and its affiliates use to develop or deploy their generative AI models and services. The services it has in mind range from voice assistants and shopping and content recommendations to enterprise solutions and developer tools. It is written as an account of Amazon's practices. The page itself does not state an effective or last-updated date.
Supporting quotations
Generative AI Development Disclosure
describes the data Amazon.com and its affiliates (collectively, “Amazon” or “we”) use to develop or deploy our generative AI models and services (“generative AI services” or “services”)
Our generative AI services range from voice assistants and shopping and content recommendations to enterprise solutions and developer tools.
Qualifications & exceptions
Covers Amazon.com and its affiliates, taken together as 'Amazon' or 'we'.
The range of services is described in general terms, not as a list of named products.
Not stated here
No effective or last-updated date appears on the page.
The page does not say which specific Amazon products count as generative AI services.
Use & Processing · Company practice
Services may run on Amazon or third-party models
Amazon says its generative AI services may be powered by foundation models, and those can be built by Amazon or by third parties. It may use several models and may choose among them to optimize performance, pick the best model for a task, or add the latest capabilities. Because the disclosure is about data Amazon uses, it does not describe what data third parties used to build any third-party model.
Supporting quotations
Our generative AI services may be powered by foundation models, which includes models built by Amazon as well as by third-parties.
We may use multiple models, and we may select models to optimize performance, select the best model for the relevant task, and incorporate the latest capabilities.
Qualifications & exceptions
Stated with 'may': this is a possibility, not a statement that every service uses third-party models.
Model choice can vary by task.
Not stated here
Which services use Amazon-built models and which use third-party models is not stated.
The training data of third-party-built models is not described here.
Data Access & Collection · Company practice
Training and testing data come from many sources, including web crawling
For the generative AI services it makes available to customers, Amazon says it trains and tests on a range of data meant to enhance the services' capabilities. The sources it says it may use are licensed and proprietary datasets, synthetic datasets, open-source datasets, and publicly available content, which expressly includes web-crawled data. These are listed as sources Amazon may use, not as a confirmed inventory for each service.
Supporting quotations
For the generative AI services we make available to our customers, we train and test on a range of data intended to enhance our services’ capabilities.
We may train and test on licensed and proprietary datasets, synthetic datasets, open-source datasets, and publicly available content (including web crawled data).
Qualifications & exceptions
Applies to generative AI services Amazon makes available to its customers.
Every source is introduced with 'may', so the list shows what Amazon may use, not what each service actually used.
Not stated here
Which sources go into which service is not stated.
'Proprietary datasets' is not defined, so this page does not say whether it includes data from customers' use of Amazon products.
Which websites are crawled, and whether site owners can exclude their content, is not described.
Content & IP Rights · Company practice
Datasets can hold many data types, including rights-protected material
Amazon says the training and testing datasets may include text, images, audio, video, code and other data types relevant to a service's purpose. It also says the datasets may contain public domain content and rights-protected material. The page does not explain on what basis rights-protected material is used, and says nothing on whether rights holders can opt out.
Supporting quotations
These datasets may include text, images, audio, video, code, and other types of data relevant to the service’s purpose.
These datasets may contain public domain content, rights-protected material,
Qualifications & exceptions
Stated with 'may': these are kinds of data the datasets can include.
Data types are tied to what is 'relevant to the service's purpose'.
Not stated here
The page gives no legal basis or permission terms for using rights-protected material.
It does not say whether creators or rights holders have any way to object or opt out.
Data Access & Collection · Company practice
Datasets may contain personal or aggregate consumer information
Amazon says the datasets may contain, 'in some cases', personal information or aggregate consumer information. The same page says Amazon applies training-data practices meant to help limit the impact of any processing of personal information in connection with training (see the safeguards card). It does not say which datasets, services or people's data are involved.
Supporting quotations
and in some cases, personal information or aggregate consumer information.
including techniques to help limit the impact of any processing of personal information in connection with training generative AI services
Qualifications & exceptions
Limited to 'some cases': personal information is not described as present in every dataset.
The safeguard language says 'help limit the impact', not that personal information is removed or excluded.
Not stated here
The page does not say how often personal information appears, or in which datasets.
It does not say whether customers' own account data or content is used.
'Aggregate consumer information' is not defined.
How Amazon collects and uses personal information is deferred to the Amazon Privacy Notice, which this analysis does not cover.
Use & Processing · Company practice
Why Amazon trains and selects models
Amazon says it trains and selects the models behind its generative AI services to deliver more accurate, helpful and relevant responses and to support service features. Its examples are responding to natural language queries, recognizing visual content, generating recommendations, and creating content. The examples are introduced with 'such as', so they are illustrations; the page does not set a closed list of purposes or promise that the data will be used only for these.
Supporting quotations
We train and select the models that power our generative AI services to help deliver more accurate, helpful, and relevant responses, and to help support the features and functionality of the service, such as by responding to natural language queries, recognizing visual content, generating relevant recommendations, or creating useful content.
Qualifications & exceptions
The examples are illustrative ('such as'), not an exhaustive purpose list.
Phrased as 'to help deliver' and 'to help support', not as guarantees of outcome.
Not stated here
The page does not say whether training data may also be used for any purpose beyond these examples.
Use & Processing · Company practice
How Amazon curates training data
Amazon says it curates training data with techniques that may include human and automated annotation, automated quality indicators, and preference ranking, among other methods. As one example it gives training data deduplication, which removes repetitive data that could cause models to overweight certain patterns or reproduce specific content. All of this is framed as what Amazon may do.
Supporting quotations
We use various techniques to curate training data, which may include human and automated annotation, automated quality indicators, preference ranking, and other methods.
we may use processes like training data deduplication to remove repetitive data that could cause models to overweight certain patterns or reproduce specific content
Qualifications & exceptions
Techniques are listed with 'may include' and 'other methods', so the list is not complete or mandatory.
Deduplication is given as an example ('may use processes like'), not a promise that it is always applied.
Not stated here
Which curation methods apply to which datasets or services is not stated.
Security & Safeguards · Company practice
Safeguards meant to limit the impact of processing personal information
Amazon says it implements multiple safeguards across its training data practices, including techniques to help limit the impact of any processing of personal information in connection with training generative AI services. The wording is that these techniques 'help limit' that impact. It does not say personal information is excluded, removed or never used, and it does not name the techniques beyond the curation examples nearby.
Supporting quotations
We also implement multiple safeguards throughout our training data practices, including techniques to help limit the impact of any processing of personal information in connection with training generative AI services.
Qualifications & exceptions
Phrased as 'help limit the impact', which is weaker than preventing or excluding personal-information processing.
'Multiple safeguards' is general; only the curation techniques in the same paragraph are named.
Not stated here
The specific safeguards for personal information are not named.
The page gives no way to check how effective the safeguards are.
Data Access & Collection · Company practice
Data volumes vary, collection began before 2022 and continues
Amazon says the amount of training and testing data varies by model or service and could range from thousands to trillions of data points. It has been collecting data since before 2022, with different models starting at different times, and it describes data collection, training and testing as ongoing so it can keep improving services and add capabilities.
Supporting quotations
The size of our training and testing data varies by model or service, and could range from thousands to trillions of data points.
We have been collecting data since before 2022, with different models beginning development at different times.
Data collection, training, and testing are ongoing processes as we continuously improve our services and incorporate new capabilities.
Qualifications & exceptions
'Could range' makes the scale a range of possibilities.
Ongoing collection is tied to continuing improvement of services and adding new capabilities.
'Since before 2022' is an approximate start, not a precise date.
Not stated here
The page does not say how long collected data are kept.
It does not say which data collection began before 2022 or for which models.
Use & Processing · Company practice
How Amazon tests and evaluates its services
Amazon says it tests and evaluates its generative AI services to assess that they meet its quality standards and perform as intended, looking at performance, accuracy and reliability for intended uses. Its methods may include automated and human evaluation, benchmarking against industry standards, simulating real-world usage and edge cases, and evaluating outputs across conditions. It also says it evaluates performance through monitoring metrics, incorporating user feedback and periodic assessments as appropriate for the service.
Supporting quotations
We test and evaluate our generative AI services to assess that they meet our quality standards and perform as intended.
Our methodologies may include automated and human evaluation, benchmarking against established industry standards, simulating real-world usage patterns and edge cases, and evaluating outputs across various conditions and contexts.
We evaluate generative AI service performance through various methods, such as monitoring metrics, incorporating user feedback, and conducting periodic assessments as appropriate for the service.
Qualifications & exceptions
Methods are listed with 'may include' and 'such as', so no particular method is guaranteed.
Periodic assessments are 'as appropriate for the service', so frequency varies.
Not stated here
The page does not say how user feedback is collected or whether it is tied to individual accounts.
It does not say what quality standards Amazon applies.
Security & Safeguards · Company practice
Nova Premier safety evaluation is the one named example
As an example of its evaluation work, Amazon says that for Amazon Nova Premier it ran comprehensive safety evaluations before deployment, including expert red teaming across critical risk domains. It names chemical, biological, radiological and nuclear capabilities, offensive cyber operations, and automated AI research and development, and says it engaged both internal experts and independent third-party evaluators. This is described for one model; the page does not say that every service gets the same evaluation.
Supporting quotations
For example, for Amazon Nova Premier, we conducted comprehensive safety evaluations including expert red teaming across critical risk domains such as Chemical, Biological, Radiological & Nuclear (CBRN) capabilities, offensive cyber operations, and automated AI research and development, engaging both internal experts and independent third-party evaluators to identify potential risks before deployment.
Qualifications & exceptions
Described as an example for one model, Amazon Nova Premier.
The evaluations are described as done to identify risks before deployment.
Not stated here
The page does not say whether other models or services received the same level of safety evaluation.
It does not report the results of the evaluations.
Security & Safeguards · Company practice
Output filtering and safety controls
Amazon says that as part of its quality evaluation process it also implements appropriate safeguards for its generative AI services, which may include output filtering or safety controls designed to enable trustworthy responses. It adds that the approaches are tailored to each service's purpose and capabilities, so which safeguards a given service has can differ.
Supporting quotations
These safeguards may include output filtering or safety controls designed to enable our generative AI services to provide trustworthy responses.
Our approaches are tailored to each service’s purpose and capabilities.
Qualifications & exceptions
Safeguards 'may include' filtering or controls; none is guaranteed for every service.
'Designed to enable' trustworthy responses describes intent, not a guaranteed result.
Approaches differ by service.
Not stated here
The page does not say which safeguards apply to which services.
Security & Safeguards · Company commitment
Commitment to responsible AI and industry collaboration
Amazon says it is committed to building AI responsibly, with appropriate safeguards for safety, accuracy, privacy and security. It also says it is working to advance responsible AI through ongoing investment in AI safety research, participation in industry standards development, and collaboration with industry partners, governments, academic institutions and safety organizations. These are statements of commitment and effort, not specific obligations toward users.
Supporting quotations
We are committed to building AI responsibly, with appropriate safeguards for safety, accuracy, privacy, and security.
This includes ongoing investment in AI safety research, participation in industry standards development, and collaboration with industry partners, governments, academic institutions, and safety organizations to advance the field of responsible AI.
Qualifications & exceptions
Framed as being 'committed to' and 'working to', which states intent and direction.
Safeguards are 'appropriate', a standard the page does not define.
Not stated here
The page does not say how these commitments are measured or enforced.
Legal Rights & Agreement Rules · Context
Other pages hold more detail on AI approach and personal information
Amazon points readers to other pages for more information: its Responsible AI at Amazon page for its approach to responsible AI, the applicable help pages for specific generative AI services and features, and the Amazon Privacy Notice for how Amazon collects and uses personal information. Those pages were not supplied with this document, so what they say is not covered here. In particular, the details of how personal information is collected and used sit outside this page.
Supporting quotations
For more information about our approach to responsible AI, see our [Responsible AI at Amazon](https://www.aboutamazon.com/what-we-do/artificial-intelligence-ai/responsible-ai) page.
For more information about specific generative AI services and features, please see the applicable help pages.
For more information about how Amazon collects and uses personal information, please see the [Amazon Privacy Notice](https://www.amazon.com/gp/help/customer/display.html?nodeId=GX7NJQ4ZB8MHFRNJ).
Qualifications & exceptions
The pointers are references only; this analysis does not read or incorporate the linked pages.
Not stated here
The content of the Responsible AI page, the service help pages and the Privacy Notice is not supplied.
The page does not say which help page applies to which service.