EBU and NVIDIA Expand Sovereign AI Framework for European Public Broadcasters

The European Broadcasting Union and NVIDIA announced at a creative industry event in Berlin that they are expanding their sovereign framework for public service media. The initiative now incorporates sovereign cloud providers, open model developers, and systems integrators to help broadcasters deploy AI agents while retaining control over data and security.

This collaboration aims to define reference architectures for where workloads run, how data is governed, and which models power services. The goal is to protect editorial independence and national resilience as AI agents move into newsrooms and production environments.

Sovereign Infrastructure and Open Models

The expanded initiative turns previous principles into deployable services by mapping standards to specific infrastructure options. Microsoft is contributing expertise from its Sovereign Cloud offerings, including Azure Local and Foundry Local. This allows broadcasters to process sensitive editorial material locally and maintain critical services even when external connectivity fails.

Nebius is providing a platform that enables broadcasters to train, fine-tune, and run their own models on European infrastructure. This ensures that the computational work remains within the region, aligning with data sovereignty requirements for public service organizations.

Open models provide the visibility necessary for broadcasters to inspect, evaluate, and adapt AI for local languages and editorial policies. NVIDIA Nemotron open models serve as a customizable foundation for domain-specific applications, allowing public service requirements to be met without relying on opaque proprietary systems.

Accelerated Media Search and Production

Adobe is also leveraging NVIDIA accelerated computing to enhance its Frame.io platform for faster video data search. The GPU-accelerated search pipeline utilizes NVIDIA NVDEC for video decoding, TensorRT for optimized AI inference, and the CUDA platform for preprocessing and embedding generation.

These technical enhancements result in approximately five times faster embedding generation, up to four times higher throughput, and up to five times lower latency for search processing. Adobe is also advancing media ingest using NVIDIA NVENC and NVDEC to accelerate encoding and decoding for growing media libraries.

A new two-frame-per-second proxy, which is about 80 percent smaller than the 360p version it replaces, reduces storage and transfer needs while retaining necessary visual detail. UFA, a leading German film and TV producer, is using these capabilities to find specific moments in video content more efficiently.

What should you do?

Organizations deploying AI agents in sensitive environments should review their data governance frameworks to ensure alignment with sovereign or local regulatory requirements. Media companies should evaluate whether current cloud infrastructure allows for local processing of sensitive editorial material during connectivity outages.

What to do and how to stay safe: NVIDIA

• Audit current AI workflows to identify where sensitive data is processed and ensure it remains within required geographic or sovereign boundaries.
• Evaluate the use of open models to allow for inspection and customization of AI behavior according to specific editorial or security policies.
• Test local processing capabilities to ensure critical media services can operate independently if external cloud connectivity is disrupted.
• Review vendor partnerships to confirm that systems integrators can deliver interoperable and practical reference architectures for your specific broadcast environment.

Step-by-step guide: Mobile Device Security: How to Protect Data Beyond Passwords

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Which companies are involved in the expanded sovereign framework for European broadcasters?

The initiative involves the EBU, NVIDIA, Microsoft, Nebius, Qvest, and other systems integrators.

How does Adobe Frame.io improve video search speed?

It uses NVIDIA accelerated computing, including NVDEC, TensorRT, and CUDA, to achieve up to five times faster embedding generation.

What is the purpose of the new two-frame-per-second proxy in Adobe Frame.io?

The proxy is about 80 percent smaller than the previous 360p version, cutting storage and transfer costs while keeping visual detail for search.