For enterprises which are prudently seeking ways to reduce their AI risk exposure, it is crucial that the solutions that they select do not themselves create additional data risks. Naturally, many companies have concerns around sharing proprietary models and private customer data contained within training and evaluation datasets. We’ve learned that there is simply no one-size-fits-all solution to handle the range of customer preferences for data privacy and ease of deployment across different verticals. Tackling this need head-on, Robust Intelligence is committed to offering a secure cloud deployment, no matter the scope of an enterprise’s data privacy needs.
Robust Intelligence offers a SaaS solution that provides companies the choice of maintaining data and compute control by either deploying a Robust Intelligence agent in their own virtual private cloud (VPC) or choosing to use an RI-managed cloud environment. Companies that prefer to keep their data on-premise can opt for the split architecture, where the customer’s models and data stay within their own cluster or VPC, while we host a control plane in an RI-managed cloud. The connection between the two environments is egress only from the data plane, meaning a customer’s cluster is never exposed to any external requests.
This solution also significantly reduces setup time when compared to an on-premise deployment, allowing customer administrators to deploy an agent within minutes and ensuring that companies can quickly begin validating their models. Additionally, administrators have the option to add multiple agents across private workspaces, or even to deploy different agents in multiple customer-side clusters, for even finer granularity of data access and compute control.
Regardless of a deployment’s agent configuration, users can rest assured that Robust Intelligence’s cloud services are held to the highest level of data compliance, as validated by our recent SOC 2 Type II report. Furthermore, across our organization, we ensure that all docker images are scanned for common vulnerabilities, invest in intrusion detection for our clusters with AWS GuardDuty, and organize regular penetration testing. Within our product, we’ve invested in the following key features to further bolster the security of our deployments.
- Encryption at rest and in transit - Whether data is in storage or in transit within the cloud, we employ AES and TLS encryption to ensure end-to-end security. TLS can also be configured within a cluster with automatically rotating certificates, for customers seeking a higher level of network protection.
- Secret Manager - All sensitive customer keys and API tokens are securely encrypted in a vault.
- SSO Authentication - Users can choose to authenticate through OIDC providers.
- Role-Based Access Control - Access to data and models within an organization can be configured as a hub-and-spoke model.
Ensuring the highest level of security for our customers is a continuous effort, and we continue to expand our capabilities to tackle new risks and match evolving standards.
At Robust Intelligence, we believe that customers shouldn’t have to compromise between speed of deployment and data security, nor should these factors prevent them from eliminating dangerous AI risks from their systems. While we also offer a comprehensive on-premise deployment option, our uniquely flexible cloud solution provides customers with both the benefits of speed and convenience of the cloud, and the assurance that their data is being handled in a safe and secure manner.
To learn more, request a product demo here.