Architecture and Infrastructure

What platform is used for bigAI?

All uses of bigAI are hosted on Amazon Bedrock . This is an AWS service that packages many different AI models into one API. Bigeye is hosted in AWS, and Bigeye’s data is stored in the same AWS environment used to access AI through Bedrock.

Both Bigeye and Amazon Bedrock have their ISO 27001 certifications and SOC 2 Type 2 attestations.

Will any data leave the AWS region where the Bigeye application is hosted?

No data will leave the AWS region where Bigeye is hosted.

Is my data sent to any third party?

No. Because Bigeye data (including all customer data stored in Bigeye’s database) is hosted in the same AWS environment being used for Bedrock, customer data is never sent over the open internet.

bigAI models used are hosted within AWS itself. AWS Bedrock has a “deep copy” of the models (e.g. if using Claude from Anthropic, there is a unique instance of the model that is only served to Bigeye) and runs the models within Bigeye’s AWS environment. It does not send requests to the model providers themselves, or anywhere else outside of the AWS environment. For more information, you can see the AWS Bedrock documentation .

Can I use bigAI if I am self-hosting Bigeye?

Yes but additional configuration is required. Please reach out to your Customer Success Engineering representative to learn more.

Controlling how bigAI works

How can I control how my data is handled by Bigeye?

Data-Restricted mode is an existing setting within all Bigeye workspaces that is used to control the behavior of all data-sensitive features, including bigAI.

When Data-Restricted Mode is off: this is the default mode for Bigeye workspaces and allows certain features to display row level data inside the user interface. In this mode, bigAI can analyze row-level data to provide additional insights, but an additional setting called “Allow bigAI to Use Row-level Data” must also be toggled on (it is off by default).

In Data-Restricted Mode: this mode blocks all access to row-level data within the Bigeye interface, including any bigAI insights. In this mode bigAI’s agent has no access to your row-level data. It will generate summaries and resolution steps using the remaining available metadata but will not be able to suggest anomalous patterns, identify subsections of impacted data, etc.

What if I have the two above row-level features on, but some users don’t have permissions to see row-level data?

bigAI will generate two sets of outputs: enhanced and reduced.

Enhanced descriptions, incident resolution suggestions, etc. incorporate insights from analyzing row-level data to identify patterns and noteworthy outliers. Reduced descriptions, incident resolutions, etc. only utilize metadata and do not have access to any row-level information when they are generated.

Users with roles which have permission to see row-level data will see the enhanced outputs. Users who do not have permission to row-level data will see the reduced outputs.

If I have the row-level features on, will my row-level data be stored anywhere?

No. Your row-level data will be queried, passed into the bigAI models to look for problematic features, e.g. a bad dimension, and then dismissed.

Bigeye does not persist the raw rows which we process and neither does AWS Bedrock. The only thing that does get persisted are attributes of the data which are found useful for debugging, for example if a particular attribute is disproportionately contributing to a data quality issue.

Will enhanced-mode outputs leak outside of Bigeye, like in alert notifications?

No. Enhanced outputs will only be displayed within the Bigeye interface, never in notifications or other forms of egress.

How your data is used

Is my data used to train AI models?

No. The Bedrock documentation states that AWS does not train models based on any customer prompts, because it receives only the “deep copy” of the models, which have been fully trained previously. AWS itself does not store Bedrock inputs or outputs. Additionally, model providers (e.g. Anthropic) do not have access to this data either.

Does Bigeye use my data for training?

No. All bigAI features within Bigeye use models that are not trained on specific customer data.

Is my data used to inform generative AI for other Bigeye customers?

No. Each call to bigAI only contains the specific information about the current context.

For example: an AI overview of an issue is created using only the information about that specific issue. And because Bigeye does not use any fine-tuned models, no metadata is used to inform any other uses of generative AI.