AI wipes out entire Swedish TV show before its premiere

[The Financial News] An unprecedented incident in Sweden saw an entire TV program deleted after an AI agent was given access to a server. While large numbers of files from personal projects have been deleted before, the final edited version of a nationally prominent TV entertainment show has now been wiped out, as the worst-case scenario involving AI becomes reality.
On the 15th (local time), according to foreign media, including Swedish daily Aftonbladet, Mastiff, the production company behind Sweden’s major TV show The Challenge Sverige, suffered the loss of the entire completed program after allowing an AI agent access to its server while putting the finishing touches on the show ahead of its premiere on November 22. The worst-case scenario became reality: the AI agent deleted the entire finished program, and the deleted files are reportedly unrecoverable.
As a result of the incident, the premiere has been abruptly postponed to 2027, and it remains unclear exactly when the show will resume airing.
Aftonbladet reported that “the work painstakingly carried out by hundreds of crew members over hundreds of days vanished in an instant,” adding that “hundreds of staff will have to be brought back in to remake and deliver the program.”
The broadcaster TV4, however, has only confirmed in an official statement that a “technical glitch” caused broadcast files to be deleted, and has avoided directly addressing whether an AI agent caused the major incident.
In an interview with the media, show host Ahmed Berhan expressed disbelief at the suggestion that an AI agent might be to blame, saying in surprise, “Really? I knew AI would kill me, but this is only the beginning.”
Experts say the incident starkly illustrates the risks of recklessly deploying unverified AI agents on critical data assets.
Recently, Anthropic’s AI coding tool Claude Code deleted nearly 50,000 files in just 103 seconds while operating under a developer’s instructions. Other problems have also been reported, including AI agents hacking government systems or leaking personal information.
[email protected] Yoon Jae-jun Reporter