A former athletic director at Pikesville High, Dazhon Darien, was arrested on Thursday for allegedly employing AI to create a false audio clip of the school’s principal. The clip, falsely attributed to the principal Eric Eiswert, was said to contain racist and antisemitic comments directed towards staff and students. This caused uproar and damage to Eiswert’s reputation leading to his temporary expulsion pending investigation into the audio clip’s authenticity.
Eiswert denied the accusations from the onset, claiming that the audio clip was a fake AI production. It provoked a strong backlash from students, teachers, and the community who were convinced that it was genuine, causing havoc in Eiswert’s life. However, AI experts after thorough analysis concluded that the audio was indeed a fake. They noted that the flat tone, absence of consistent breathing sounds or pauses, and unusually clear background sounds were indicative of it being AI-generated.
The investigation led to the arrest of Dazhon Darien on multiple charges including disrupting school operations, theft, retaliating against a witness, and stalking. It is claimed that Darien created the fraudulent audio using AI as payback after Eiswert initiated an investigation into Darien’s alleged misuse of school funds. He reportedly misused the school’s computers to access OpenAI tools and Microsoft Bing Chat services for the act.
Interestingly, there was no charge related to using AI to falsify someone’s voice, because there is not yet a legal rubric for such a crime. Proposals such as the No Fakes Act and the No AI Fraud Act have been presented in the US Congress, although they have not been passed. Thus, Darien’s act of creating a non-consensual fake audio using AI, while ethically wrong, did not constitute a legal violation at the time.
The incident underlines how society and authorities are still struggling to catch up with the numerous challenges presented by AI. The growth and sophistication of generative AI present new opportunities for misuse, as demonstrated by the fact that even a school teacher was able to produce a rather convincing fake. This incident brings to fore the inherent danger and the need for better mechanisms to identify fake audio and video. Until more robust authentication methods are in place, the pragmatic approach is to treat any suspicious content as potentially fake until verified as real.
