OFF Radio Kraków Dropped Its AI Hosts After One Week
OFF Radio Kraków’s AI-presenter experiment remains discontinued. The Polish channel ended the format after one week in October 2024, and its current editorial roster identifies a newsroom manager, a music editor and nine human collaborators.
The episode is therefore not an ongoing replacement of an entire station’s workforce. It was a brief experiment that collapsed as a programming model amid a much larger argument about creative labor and synthetic identity: the Associated Press account of the October 2024 shutdown records the one-week run, a protest petition with more than 23,000 signatures and the broadcast of a simulated interview with Wisława Szymborska.
What the station put on air
OFF was a youth-oriented thematic channel operated by Poland’s public broadcaster Radio Kraków. Its experimental schedule featured three fictional presenters whose biographies, portraits and voices were created with generative tools. They were designed as young characters covering culture, society and subjects intended to interest Generation Z.
The presenters were not autonomous systems controlling the broadcast. Human journalists developed their character profiles, prepared or checked the material and converted approved text into synthetic speech. Music selection also incorporated suggestions from AI tools.
That production chain makes “AI-run station” an imprecise description. The audience encountered fabricated on-air personalities, but people still exercised editorial control behind them. The experiment replaced the visible and audible role of a presenter more directly than it replaced every journalistic function.
Why audiences saw replacement, not experimentation
The controversy was rooted in timing. OFF’s previous contributors had lost their roles before the channel returned with virtual personalities, so the public saw recognizable human hosts disappear and synthetic substitutes occupy the same position. Former contributor Mateusz Demski’s petition framed the change as a precedent that could threaten employment throughout journalism and other creative industries.
Management treated the staffing decisions and the AI project as separate matters. Its position was that the former format had attracted too small an audience and that the relaunch was intended to examine the opportunities and risks of artificial intelligence in media.
The available public record does not independently establish that AI caused the earlier contracts to end. It does establish a communications failure: the distinction between a contractual decision and an editorial experiment could not outweigh the substitution listeners could see and hear.
That distinction remains useful for the creator economy. Introducing AI backstage—for transcription, research assistance or production cleanup—does not necessarily alter the identity through which an outlet builds trust. Replacing the person who appears to speak changes authorship, accountability and the audience relationship, even when humans continue working elsewhere in the process.
The Szymborska simulation changed the stakes
The project moved beyond fictional hosts when an AI presenter appeared to interview a synthetic version of Wisława Szymborska. The Nobel Prize’s official record identifies Szymborska as the 1996 literature laureate and gives February 1, 2012, as the date of her death.
The station labelled the segment as AI-produced and indicated that the president of the foundation responsible for the poet’s legacy had consented to the project. Those safeguards addressed disclosure and authorization, but they did not turn generated remarks into authentic testimony.
An interview normally implies that a subject heard questions and chose the answers. A posthumous simulation cannot provide that exchange, so the format borrowed the authority of journalism for material that was necessarily invented. Clear labelling reduced the risk of literal deception, yet it could not resolve the more fundamental question of whether an artificial reconstruction should use the conventions of a genuine interview.
This was a different issue from using fictional presenters. A made-up host requires visible disclosure and named human accountability; a simulation of a real person also introduces consent, likeness and attribution concerns. When the person is dead, the publisher must additionally prevent generated language from being mistaken for something the subject actually believed or said.
Why the project ended so quickly
The AI-led programming was initially expected to remain a temporary experiment, with a possible duration of up to three months. It stopped after a week, and OFF returned to a mainly music-based format distributed online and through digital radio.
Management characterized the intense response as evidence that the project had already produced enough observations and debate. Critics could assess the same outcome by a different standard: the employment dispute overwhelmed the programming, while the Szymborska segment blurred the boundary between disclosed simulation and documentary speech.
Both interpretations can coexist because they measure different things. The project succeeded at attracting attention to AI in media, but it failed to establish virtual presenters as a durable format. Public controversy is not evidence of repeat listening, audience loyalty or improved trust.
No publicly documented audience methodology attached to the experiment demonstrates lasting listener growth attributable to the virtual hosts. Claims that curiosity produced a particular audience surge should therefore not be treated as proof that synthetic personalities made the channel commercially or editorially stronger.
What the one-week reversal actually demonstrates
The clearest conclusion is narrower than either “AI replaced radio journalists” or “the experiment achieved its goal.” Generative tools were capable of producing usable voices, faces and scripts, but the resulting format did not earn enough legitimacy to survive beyond its first week.
Disclosure was necessary but insufficient. Listeners were not judging only whether the content had been labelled; they were also judging who had lost visible work, who was responsible for generated speech and whether the format respected the difference between a fictional character and a real historical person.
The channel’s subsequent operation with publicly named human contributors provides the most concrete update. OFF Radio Kraków did not adopt its three virtual presenters as a permanent substitute for an editorial team. The lasting significance of the episode is the speed with which questions of labor, authorship and trust eclipsed the technology itself.
Also read:
Subscribe to our newsletter
Get the latest Web3, AI, and crypto news delivered straight to your inbox.