Deepgram, in collaboration with Fortanix and NVIDIA, has announced a significant upgrade to its voice AI technology, which now operates fully on-premises with enhanced privacy features. This new deployment allows organizations, particularly those in regulated sectors like healthcare and finance, to utilize real-time voice AI while ensuring that sensitive audio data and AI model weights remain protected during active processing. This advancement leverages confidential computing and a pre-integrated stack that keeps data encrypted and secure in trusted execution environments, addressing the growing compliance needs under regulations such as HIPAA and GDPR.
NVIDIA: NVIDIA develops GPUs with Confidential Computing features that create trusted execution environments for AI workloads, isolating processing from the underlying infrastructure. The partnership leverages these capabilities alongside Fortanix to deliver hardware-rooted security for Deepgram’s voice AI models in sensitive enterprise settings.
Deepgram: Deepgram is the real-time AI infrastructure company underpinning the Voice AI economy, providing speech-to-text, text-to-speech, and speech-to-speech capabilities through enterprise-grade models. In this news, it announced a partnership to enable secure, self-hosted voice AI deployments for regulated industries using confidential computing technology.
Fortanix: Fortanix is the global leader in data and AI security and a pioneer of Confidential Computing, offering a unified platform to protect sensitive data, AI models, and applications across environments. It is central to the announced partnership, supplying technology that protects model weights and audio data during active inference in on-premises setups.
Anand Kashyap: Anand Kashyap is the CEO and co-founder of Fortanix. He commented on the partnership’s role in helping AI model owners commercialize their IP securely for enterprise customers without exposing it to risks.
Justin Boitano: Justin Boitano is vice president of Enterprise AI Platforms at NVIDIA. He highlighted the solution’s value for industries like healthcare, where real-time voice AI can scale efficiency while maintaining security throughout data processing.
Scott Stephenson: Scott Stephenson is the CEO and Co-Founder of Deepgram. He is quoted in the announcement emphasizing how the partnership enables organizations to run voice AI without compromising data or model protection in highly regulated environments.
`json
{
“Security”: “Confidential computing ensures AI models and data stay encrypted and secure during active processing within trusted execution environments.”,
“Deployment”: “The collaboration provides a pre-integrated solution combining voice AI models with hardware-isolated settings, facilitating secure, enterprise-grade voice applications for organizations with stringent data sovereignty requirements.”,
“Regulation”: “Industries like healthcare and finance require on-premises voice AI solutions that secure sensitive audio and proprietary models to comply with regulations such as HIPAA and GDPR.”
}
`