"Human in the Loop"

Private broadcasters adopt AI guidelines

Austria's private broadcasters have drawn up guidelines for dealing with artificial intelligence. According to a press release from the Association of Austrian Private Broadcasters (VÖP), they are intended to provide orientation and demonstrate the trustworthy use of AI. 

The guidelines stipulate, for example, that content created with the support of generative AI must always be checked by employees before publication ("human-in-the-loop"). The guidelines are intended to supplement the European AI law, whose obligations will only gradually come into force over the next three years. In principle, the broadcasters are committed to an open and critical approach to AI.

AI can simplify work and improve results. However, it is clear that the use of AI in programming poses particular risks in terms of content quality and user trust. AI tools should therefore be thoroughly tested before they are used. AI-supported format conversions - such as from speech to text - are excluded from human pre-testing.

In terms of transparency, the guideline states that the use of generative AI in particular will be indicated if this is necessary for journalistic reasons. If voice, facial expressions, gestures or similar data from people are used to train AI models, there is an awareness of the special responsibility that this entails.