ElevenLabs and Deutsche Telekom Introduce AI Calling Through Network Infrastructure.

ElevenLabs and Deutsche Telekom introduced AI Calling at Mobile World Congress, a system that embeds AI translation and support directly into phone calls through network infrastructure. The feature will launch in 2026 and aims to provide real-time assistance without requiring a separate app or device.

The announcement marks a shift from app-based voice AI tools to network-level integration. Instead of downloading software, users will access AI translation, transcription, and conversational assistance as part of the phone call itself, delivered through Deutsche Telekom's network.

AI Calling is designed to translate conversations, provide contextual support, and help callers understand one another during live calls. The system operates through the carrier network, meaning the AI layer runs server-side rather than on the user's device.

According to the announcement, the technology will be available directly inside the call, positioning it as infrastructure rather than a consumer product. Deutsche Telekom's network will handle the processing, making the feature accessible to any subscriber without additional hardware or app installation.

This approach contrasts with existing voice AI services from ElevenLabs, which typically require API integration or app-based deployment. The company has recently expanded its voice agent capabilities for customer support use cases, but AI Calling targets a different layer: the telecommunications network itself.

The announcement invoked historical context, referencing Albert Einstein's 1924 radio broadcast from Vox-Haus in Berlin, now the site of Deutsche Telekom's headquarters. The parallel positions AI Calling as the next evolution of broadcast communication technology, applying intelligence to an infrastructure that has remained largely unchanged for decades.

While ElevenLabs and Deutsche Telekom did not disclose specific implementation details, network-level AI deployment introduces challenges around latency, privacy, and interoperability. Real-time translation during a phone call requires low-latency processing to avoid conversational delays, meaning the AI models must run efficiently on carrier infrastructure.

The system will need to handle multiple languages simultaneously, manage cross-border calls with varying network conditions, and maintain call quality while processing audio streams in both directions. Deutsche Telekom's role suggests the feature may initially roll out to its subscriber base before expanding to other carriers.

ElevenLabs has been developing multilingual voice models for dubbing and speech synthesis, including Dubbing v2, which supports over 30 languages. AI Calling appears to apply similar speech technology to synchronous, two-way conversations rather than pre-recorded content.

The partnership is part of a broader trend toward embedding AI at the infrastructure level. Telecommunications providers are under pressure to differentiate their services beyond data and voice plans, and AI-powered features offer a potential path to added value. For ElevenLabs, the deal represents a move beyond APIs and developer tools into carrier-grade deployment.

ElevenLabs has opened a sign-up page for users interested in early access to AI Calling. The feature is expected to launch within the next year, though the company has not specified whether the rollout will be limited to Germany, available across Deutsche Telekom's European footprint, or expanded globally through partnerships with other carriers.

No pricing model was disclosed, leaving open the question of whether AI Calling will be bundled into existing mobile plans, offered as a premium tier, or charged on a per-use basis. The network-level implementation suggests it could be carrier-subsidized, but business models for AI-enhanced telecommunications remain largely unproven at scale.

The announcement positions AI Calling as a product category rather than a single feature, suggesting future iterations could include meeting transcription, voice-controlled automation, or accessibility tools for users with hearing impairments.