EBU and NVIDIA to write shared rules for AI agents at public broadcasters

The European Broadcasting Union (EBU) and NVIDIA are expanding their sovereign AI work for public service media. The goal is to define how members can deploy AI agents in newsrooms, production and audience services without giving up control of their data, infrastructure, models and security. Microsoft, Nebius and systems integrator Qvest have joined. The announcement came on Monday, October 19, at 17:00 CEST, during a creative industry event in Berlin.

According to NVIDIA, the EBU has more than 115 member organizations in over 55 countries, reaching more than a billion people. What the two describe is a shared set of requirements that could shape how Europe's public broadcasters buy and build AI services in the coming years.

From principles to reference architecture

NVIDIA and the EBU introduced their first sovereign AI standards for public service media in June 2025. The work now moves from principles to deployable services. NVIDIA says the initiative will define requirements and reference architectures covering where workloads run, how data is governed, which models power services and how systems stay secure throughout their lifecycle.

The driver is agents. As AI agents move into newsrooms and production, they gain access to unpublished material, archives and publishing systems. The EBU ties this to editorial independence, data sovereignty and national resilience.

"These standards give EBU members a shared foundation to strengthen resilience, retain control and deliver accountable advanced services that sustain audience trust," said Antonio Arcidiacono, chief technology and innovation officer at the EBU.

Who does what

Cloud providers will map the standards to deployment options. Microsoft brings experience from Microsoft Sovereign Cloud, including Azure Local and Foundry Local. According to the announcement, this lets broadcasters process sensitive editorial material locally and keep critical services running when external connectivity is down.

Nebius offers a platform for broadcasters to train, fine-tune and run their own models on European infrastructure. NVIDIA points to its open Nemotron models as a base broadcasters can inspect, evaluate and adapt to local languages, archives and editorial policies. Qvest and other systems integrators are meant to make the reference architectures work in real broadcast environments.

Delphine Ernotte, CEO of France Télévisions and president of the EBU, said sovereign cloud infrastructure and open models can help Europe's public service media innovate while protecting data, editorial independence and cultural diversity.

Adobe and UFA show the production side

In the same announcement, NVIDIA describes how Adobe uses NVIDIA GPUs to speed up video search in Frame.io. Adobe reports roughly 5x faster embedding generation, up to 4x higher throughput and up to 5x lower search-processing latency. A new two-frame-per-second proxy is about 80 percent smaller than the 360p version it replaces.

German producer UFA, which makes more than 3,500 hours of programming a year, uses the search to find specific moments in up to 15 years of footage from the daily drama "Gute Zeiten, schlechte Zeiten". Editors can search by concept, narrow by character name and filter to approved footage. This is a performance upgrade to an existing product, not part of the EBU standard.

What is missing

The announcement does not say when the reference architectures will be finished, which EBU members are taking part, or whether any broadcaster has deployed them yet. It gives no pricing for Microsoft, Nebius or NVIDIA services, and no detail on administration, access control or auditing of the agents. Whether the standards will be binding on members is also not stated.

What leaders should note

For media executives, the key point is that the EBU is now writing shared requirements for where AI agents run and how they are secured. Requirements like these tend to become procurement criteria. Anyone planning agents inside editorial systems should track the work before locking in their own architecture.

It also matters who is writing the standard. NVIDIA, Microsoft and Nebius are vendors with products at stake. A sovereignty standard shaped with suppliers may reduce dependence on individual cloud services, but it can also create new dependencies on hardware and software from the same companies. That is worth assessing once the reference architectures are published.

Sources and media

• NVIDIA Blog: NVIDIA Advances Agentic Security for Media, From European Public Service Broadcasting to Global Content Production – published October 19, 2026, 17:00 CEST.
• NVIDIA Blog: EBU and NVIDIA on sovereign AI for public service media (June 2025)

Illustration: AI-generated with Grok Bot / hogby.ai