Misuse of new technologies: how much do deepfakes affect the media?
An Ekipa article on how ready Montenegrin newsrooms are for deepfakes, with a three-level verification method and a plan for when a fake appears.
Written by journalist Višnja Cvetković for Ekipa, a portal of the Trade Union of Media of Montenegro, this article looks at how ready Montenegrin newsrooms are for fake audio and video made with artificial intelligence. It starts from a recording that raised the question of how quickly a newsroom can verify material before it spreads, and it brings together three sources: Dražen Đurašković of the Trade Union of Media of Montenegro, Damir Nikočević of the Agency for Audiovisual Media Services and Branko Džakula, cybersecurity expert and co-founder of Secfix.
Džakula’s diagnosis is that most Montenegrin newsrooms have no written procedure for checking audio and video that comes from outside, no named person responsible for it and no forensic contact arranged in advance who can respond within hours. His examples show why that matters: a fake advertisement for an investment platform that used the likeness of a news presenter to borrow the audience’s trust, a fake audio recording published in Slovakia in September 2023, two days before the election and during the pre-election silence, which made a timely denial harder, and research since 2019 consistently showing that most deepfake material online is non-consensual sexual content, the vast majority of it aimed at women.
He sets out verification in three levels: the origin of the material first, then independent confirmation, and technical analysis last. Independent confirmation means reaching the person whose face or voice is used through a channel established in advance and comparing the material with the archive, down to whether the programme aired that day and whether the opening titles, graphics, on-screen caption, set and studio clock match, and for urgent communication he recommends a control word or a question only the real source can answer. Technical analysis looks at the edges of the jaw and hair, the rhythm of blinking, lips that do not match the voice, room acoustics and breathing, and since automatic detectors become less reliable after repeated compression and re-encoding, their result should be treated only as an indication. The other two sources add the union’s warning that authentic recordings can now be dismissed as fakes and the regulator’s view that the 2024 law on audiovisual media services has no separate category for deepfakes, while the Agency has so far received no complaints about content made or altered with artificial intelligence.
Key findings
- 01Most Montenegrin newsrooms have no written procedure for checking outside audio and video, no named person responsible and no forensic contact who can respond within hours.
- 02Verification runs in three levels: the origin of the material, independent confirmation through a channel established in advance and the archive, and only then technical analysis.
- 03Automatic detectors lose reliability after repeated compression and re-encoding, so their result is only an indication.
- 04Newsrooms should keep original files and timestamps, use Content Credentials (the C2PA standard) and maintain a public archive of their broadcasts.
- 05A response plan needs a prepared denial, a reaction within one hour, a named spokesperson, takedown reports prepared for the platforms and contact with Montenegro’s national CIRT and the police, with links, screenshots, timestamps and file hashes collected at once.
- 06Newsrooms should run at least two practical trainings a year, including a simulated incident with a fake made from their own material, and treat an attack on a journalist’s face or voice as an attack on the newsroom itself.