Deepgram and Fortanix today announced a partnership that will enable enterprises to run voice AI in their own environment on their own terms while ensuring their most sensitive data is securely protected. Under terms of the agreement, Deepgram can leverage Fortanix Confidential AI and NVIDIA Confidential Computing to add an additional layer of advanced security to self-hosted environments to ensure that its proprietary model weights, built on business-critical intellectual property, can be deployed while protecting against model theft or inappropriate use. With this announcement, Deepgram and Fortanix continue to raise the bar for model-in-use protection in the most security-sensitive on-prem environments, enabling increased voice AI adoption in highly regulated industries.
For enterprises, especially those in highly regulated industries, security requirements continue to tighten. Organizations handling patient conversations, financial transactions, or classified information increasingly require that sensitive audio and AI model weights remain protected not only at rest and in transit, but also during active processing in their own environments. This level of protection enables organizations to build highly-secure real-time voice applications without sacrificing on performance.
The on-premises solution runs Deepgram’s voice AI models with Fortanix Confidential AI on NVIDIA Confidential Computing-enabled GPUs, creating a hardware-isolated environment where both audio data and model weights remain encrypted and protected throughout active use. NVIDIA GPUs with Confidential Computing enable AI workloads to process sensitive data inside a trusted execution environment — a capability traditional infrastructure cannot provide. By bringing together best-in-class voice AI models, hardware-rooted isolation, and a jointly engineered, pre-integrated stack, the partnership delivers a level of in-use data protection that, until now, has not been practical to deploy at enterprise scale.
The Deepgram, Fortanix, and NVIDIA solution opens the door to a variety of on-prem security-demanding voice AI applications: private, on-prem voice agents handling sensitive customer and patient interactions; enterprise-wide transcription layers that capture every call, meeting, and internal conversation for analytics, compliance, and search; and voice-enabled IT, operations, and service desk applications running entirely inside an organization’s secure perimeter. For regulated enterprises, this turns voice into a production-ready interface without sacrificing the real-time performance the experience demands.
Deepgram’s voice AI models deliver the real-time voice understanding and generation with the accuracy, consistency, and low latency that enterprise use demands. Designed for any environment including those with the highest confidentiality and regulatory needs, Deepgram’s models bring voice AI to enterprise organizations across virtually every industry vertical, including those with sensitive, regulated use cases that have historically been out of reach.
Fortanix Confidential AI protects data and AI model weights while they’re actively running. It builds on NVIDIA GPUs with Confidential Computing to create Trusted Execution Environments (TEEs) that isolate the AI workload from the underlying infrastructure and OS. Data and AI models run safely inside Confidential Computing, encrypted in memory, and inaccessible to the host operating system or even privileged administrators. As a result, regulated organizations can unlock AI innovation with trust, security, and sovereignty at the core, while meeting HIPAA, GDPR, and national-data residency requirements.
To learn more, please reach out to Deepgram at: partners@deepgram.com.
Deepgram Delivers Real-Time Voice AI at the Edge for Use with Snapdragon
Posted in Commentary with tags Deepgram on July 21, 2026 by itnerdDeepgram today announced an initiative to bring enterprise-grade speech recognition directly onto PCs powered by Snapdragon® processors. By optimizing Deepgram’s Nova-3 speech-to-text model on the Qualcomm® Hexagon™ NPU in the Snapdragon X Series platform, Deepgram is enabling developers and device manufacturers to deliver real-time voice experiences with greater speed, privacy, and reliability, without relying on a cloud connection. This effort opens the door to further integration of voice into a new generation of intelligent applications across automotive, mobile, AI PC, XR, industrial edge, IoT, and wearable devices.
Many voice AI solutions have historically relied on a cloud-based architecture. Before a response could be delivered, each interaction required audio to leave the device, travel to the cloud, be processed somewhere else, and then return. With this approach, delays and privacy concerns are sometimes introduced, which limit where voice AI can realistically be deployed. Deepgram is fundamentally changing that model, enabling speech recognition to happen directly on the device itself. The result is an entirely new world of applications and user experiences that feel like a natural conversation, whether it is running in a vehicle, on an AI PC, inside an XR headset, or at the edge of a network where connectivity cannot be guaranteed.
Nova-3 advances Deepgram’s industry-leading accuracy, extending its capabilities to a broader range of real-world enterprise use cases and challenging audio conditions. It is the first voice AI model to offer real-time multilingual transcription. Deepgram is also the first to provide users with demonstrably effective and highly accurate self-serve customization – enabling instant vocabulary adaptation without model retraining. Superior accuracy: Nova-3 leads transcription accuracy with a 6.89% word error rate on real-world production audio, a 24.7% lower error rate than the next-best competitor.
To learn more, please visit: https://deepgram.com/partners/qualcomm.
Leave a comment »