Mavenir and Sanas partner to embed Speech AI in telecom cores

Mavenir and Sanas are integrating real-time speech AI into carrier networks, adding translation, voice enhancement and deepfake detection at telecom scale.

By The Datatech Times Editorial Team
A brightly lit, modern control room features a large curved screen displaying a network graphic, multiple workstations with black monitors and chairs, and server racks on either side, with natural light entering through tall windows and ove

Mavenir, the cloud-native telecoms software vendor, has formed a strategic partnership with Sanas, a real-time speech AI platform, to bring speech enhancement, live translation and synthetic-voice fraud detection natively into mobile core networks. The integration places Sanas technology inside Mavenir's MAVcore Voice AI portfolio, enabling communication service providers to offer the capabilities as operator-grade services rather than add-on application-layer features.

The partnership is in technical validation; commercial availability for operators has not been dated. Mavenir and Sanas plan to showcase the joint solution at the AI-Native Telco Forum in Düsseldorf in September and at FutureNet Asia in Singapore later the same month.

What the integration delivers

At its core, the Sanas platform performs three functions in real time: accent and language normalisation, audio quality enhancement, and deepfake or voice-clone detection. Embedding these capabilities inside the network rather than at the application edge is significant because it removes dependence on third-party API calls and keeps processing within the operator's own infrastructure, which the companies say satisfies data-residency and sovereign AI requirements.

A notable existing deployment of the Sanas technology, cited in the announcement, is a Language Assist arrangement with MicroAutomation that enables emergency-services dispatchers in the US to hear callers in English without an interpreter. Bringing that kind of latency-sensitive, mission-critical speech processing to carrier scale is the stated ambition of the Mavenir deal.

Pardeep Kohli, president and chief executive of Mavenir, said the aim is to let operators "transform voice from a connectivity service into a monetisable, AI-powered experience" by embedding translation, voice quality and security functions directly within the network. Service configuration and monetisation are orchestrated through Mavenir's AI Agent Studio, which the company positions as a control layer for creating and pricing new conversational AI experiences.

Market context and competitive landscape

The partnership sits at the intersection of two fast-moving markets: AI-enhanced voice and telecom network softwarisation. Operators have long struggled to differentiate voice, which has commoditised as over-the-top messaging and calls eroded legacy revenues. Embedding AI at the network layer, rather than leaving it to handset manufacturers or app developers, is a strategy that several telecoms software vendors are pursuing, including cloud-native rivals such as Ericsson's cloud-native portfolio and a growing cluster of specialist AI voice companies.

On the security side, synthetic-voice fraud is an escalating threat. Fraudsters are already using voice-cloning tools to impersonate individuals in social-engineering attacks and authorisation calls. Carrier-level deepfake detection, if it can operate at the latency tolerance of a live call, would represent a meaningful defensive layer that application-level solutions currently cannot match without network co-operation.

Sanas reported strong growth credentials in its boilerplate: the company was named the top-ranked AI and data company in the United States on the 2026 Inc. 5000 list of fastest-growing private companies. The release did not disclose revenue, valuation, or the number of operator customers currently under contract.

Regulatory and standards read-across

Sovereign AI and data residency are explicit design goals of the integration. That framing is commercially important: EU operators are subject to GDPR restrictions on transatlantic data transfer, and the EU AI Act's provisions on general-purpose AI systems and high-risk applications will apply to automated voice processing used in emergency or financial contexts. Offering processing inside the operator's own network boundary provides a compliance pathway that cloud-hosted alternatives may struggle to match in regulated markets.

The GSMA's ongoing work on Open Gateway APIs and network-as-a-service frameworks is also relevant here; operators that expose voice AI capabilities through standardised APIs could build an ecosystem of third-party services on top of the Mavenir-Sanas stack, expanding the addressable revenue opportunity beyond direct-to-subscriber offerings.

Sector

Topic

Geography