Privacera AI Governance Integrates with AWS on Foundational Model Security
November 2023 by Marc Jacob
Privacera announced that Privacera AI Governance (PAIG) now integrates with Amazon Web Services (AWS) on security for foundation models (FMs) used for generative AI. PAIG is designed to provide the ability to responsibly govern and protect sensitive data within FMs and generative AI applications. PAIG does this by leveraging the power of Amazon Bedrock, a fully-managed service that makes FMs from leading AI companies accessible through an API to build and scale generative AI applications, and Amazon SageMaker, a cloud-based machine-learning platform that enables developers to create, train, and deploy machine learning (ML) models on the cloud, to support open-source and proprietary FMs and workflows. AWS services uphold enterprise-grade security and privacy best practices, and with PAIG, customers can take security and privacy measures even further.
PAIG provides a comprehensive suite of built-in product capabilities to address privacy, security, and compliance requirements associated with building generative AI applications. Whether using some of the open-source, public FMs or customizing private FMs, the same consistent security controls can be applied to training and tuning data, as well as user-injected model inputs and outputs.
The new integration provides the following security and governance capabilities covering the end-to-end lifecycle of generative AI applications - from discovery, training, and deployment, to continuous monitoring:
Privacera’s Unified Data Security Platform ensures the masking and redaction of sensitive training and tuning data, while PAIG secures the generative AI models and applications. PAIG specific capabilities allow organizations to:
Prevent Sensitive Data Leakage
• PAIG provides the ability to define governance and security policies using easy-to-build and understandable policies created using natural language and to easily enforce these policies across any generative AI application or model.
Detect and Filter for Risk and Abuse
• PAIG detects sensitive data in AI model output and input by analyzing user injected model inputs and outputs and blocking or masking data that could expose the model or model users to data they are not authorized to see.
Observability and Traceability
• PAIG monitors and analyzes user interactions with the AI models and provides dashboards that provide visibility across all generative AI applications, and models, including type of requests made, sensitive data identified, and actions taken to protect sensitive data. Privacera also provides a comprehensive audit trail to track individual user activities with detailed information of individual requests and specific security applied.
The integration with Amazon Bedrock extends the existing, comprehensive integrations Privacera offers as a comprehensive unified data security governance solution for over 20 AWS services: ranging from data and analytics services, like Amazon Athena, Amazon EMR, Amazon OpenSearch Service, Amazon Redshift, and Amazon Relational Database Services (Amazon RDS), Amazon Simple Storage Service (Amazon S3), and third-party services that run on AWS, like Databricks and Snowflake, to Privacera’s seamless integration with AWS Lake Formation.