Neck-Deep in Digital Oil? Public Broadcasters’ Archives as AI Training Datasets

The accelerating demand for artificial intelligence (AI) training datasets has brought media archives into focus, with public broadcasters’ archives – rich in audiovisual content and metadata – attracting interest from both affiliated developers and big tech. This article explores the strategic, legal, and ethical considerations surrounding their use as AI training datasets. 

Strategically, public broadcasters must weigh financial opportunities against reputational risks. While collaboration with AI companies can generate revenue and goodwill, it may erode public trust if seen as compromising journalistic integrity. Legal challenges include navigating copyright, data protection laws, and restrictions on monetizing publicly funded assets, alongside limitations imposed by existing agreements with third-party providers. 

Ethically, these archives, as repositories of societal memory and public values, pose profound dilemmas. Should they fuel AI algorithms that risk perpetuating bias or undermining democratic discourse? Balancing the imperative to ensure responsible use with the need to respect content creators’ intentions adds complexity. This presentation situates these debates within the broader context of a rapidly expanding AI training dataset market and shifting archival business models. It underscores the delicate balance between maximizing societal value and upholding the public service mission of broadcasters. While strongly rooted in policy development work for administrators of media companies, the insights of this article will also offer value to scholars in media studies, media history and policy making.