AI governance in broadcasting should begin with a practical question: what happens between someone opening a tool and its output reaching an audience?
There is a recent industry reference point. In its position paper dated 25 September 2026, the European Broadcasting Union describes AI applications across production, distribution, content management and journalism. It also highlights an AI-supported cross-border news initiative in which journalists retain control of editorial decisions. This is an industry policy position, not a new regulatory requirement.
For broadcast teams, the following measures offer a starting point.
Inventory uses, not just tools. An approved application can support very different activities: transcribing an interview, suggesting archive tags or generating a presenter’s voice. Record the purpose, accountable owner, information involved and intended audience for each use. Assess the workflow rather than assuming one approval covers everything.
Set boundaries before uploading material. Establish rules for confidential scripts, personal information, licensed footage and unreleased recordings. Review supplier terms covering retention, model training, access and deletion. Where permissions are unclear, pause the upload and seek an appropriate review.
Make human review specific. “Human oversight” needs an identifiable reviewer and a checklist. For captions and translations, that might mean checking names, numbers, meaning and timing against the original recording. For generated imagery or audio, review should address permissions, potential audience confusion and any disclosure needed.
Connect governance to asset management. Keep source references, approval status and relevant AI-use notes with the media asset where practical. Define how teams correct an error after material has moved into editing, distribution or the archive. A detached policy document cannot substitute for usable records.
Plan for change. Reassess workflows when suppliers change models or terms, and when teams expand how they use a tool. Provide a clear route for reporting unexpected output without discouraging staff from raising concerns.
The objective is to make responsibility visible and review proportionate to the risk. Start with one bounded workflow, document its checks and limitations, and expand only when the operating process is understood.
