The EBU has outlined a future distribution model in which more AI processing happens on customer-premises equipment and audience devices rather than in central data centres.
Potential uses include local recommendation, translation, lip-sync, subtitles, audio description, sign-language rendering and personalised sports overlays. The proposal also connects local processing with C2PA provenance and hybrid broadcast or peer-to-peer distribution.
Why it matters
Central AI services can create privacy, cost and energy problems because both user data and media travel repeatedly across the network. Edge execution can keep personal context local and reduce backbone traffic, while broadcast delivery remains efficient for common content.
The trade-off is operational fragmentation. Broadcasters would need to support different chip capabilities, model versions and security states in consumer devices. A credible deployment model therefore needs lifecycle management, measurable energy accounting and a fallback experience for devices that cannot run the local functions.
