Inpher SecurAI protects the privacy of user inputs on large language models
Inpher released SecurAI, a solution that protects the privacy and security of user inputs on large language models. This release of SecurAI leverages the NVIDIA H100 Tensor Core GPU for maximum speed and performance.
“Enterprises need to harness the power of generative AI while seamlessly and securely incorporating it into their business processes,” stated Daniel Rohrer, VP of Product Security at NVIDIA. “Inpher’s SecurAI coupled with NVIDIA confidential computing technology enables organizations to use confidential GPUs for improved data utility and business productivity.”
Powered by confidential GPUs, this much anticipated release of SecurAI also integrates retrieval-augmented generation (RAG). SecurAI ensures that user prompts are never exposed to the model provider while enabling organizations to improve model outputs by seamlessly integrating with core business services, data sources and applications such as Slack, Zendesk, Salesforce, Snowflake, et al. SecurAI also enables enterprise users to upload their own files based on their organizational needs by leveraging secure RAG.
“Security is a critical element for every enterprise to consider when building a generative AI deployment strategy,” stated Dr. Jordan Brandt, CEO at Inpher. “Our commitment to privacy and the responsible use of AI is foundational to our team and by working closely with NVIDIA we deliver the most robust, enterprise-ready generative AI solution.”
Inpher SecurAI addresses the privacy and security concerns that users have when trying to leverage AI and open models where prompts might contain sensitive data that needs to be protected from the model and/or cloud service provider. The intricate details of the prompts received and the subsequent data processing occur within a Trusted Execution Environment that is isolated from the underlying infrastructure.
This isolation shields sensitive information, ensuring that the specifics of user inputs, such as, proprietary business data, personal details, or other confidential information, remain inaccessible to external entities, including SecurAI’s own hosting infrastructure.
By keeping the intricate details of user inputs and data processing within a protected space, SecurAI significantly reduces the risk of data breaches, leaks, or unauthorized access, making it a strong choice for safeguarding IP and privacy. SecurAI’s distinct focus on data protection and privacy offers a secure choice for organizations emphasizing the privacy of their data and intellectual property. With SecurAI, users can incorporate large language models into their organization without compromising privacy.