On Thursday, Baltimore County Police arrested Pikesville High School’s former athletic director, Dazhon Darien, and charged him with using AI to deceive Principal Eric Eiswert, according to a report by The Baltimore Banner. According to the press, Darien allegedly used AI voice synthesis software to imitate Eiswert’s voice, giving the impression that the director made racist and derogatory remarks.
The audio clip, posted on a popular Instagram account, contained insulting remarks about “ungrateful Black kids” and their academic achievement, as well as a threat to “address the other side” if the listener received one more issue from “one more Jew in this society”. Additionally, the recording mentioned names of staff members, including Darien’s alias “DJ,” making it seem as though they should not have been hired or “to be removed” in some way or another.
The remarks led to significant controversy from students, faculty, and the wider society, many of whom initially believed the superintendent had actually made the comments. Shaena Ravenell, a professor at Pikesville High School, presumably played a significant role in the audio distribution. Police said she forwarded the contentious message to a scholar known for their ability to spread information immediately via social media, despite not being charged. Therefore, this student increased the scope of the audio, including sharing it with the press and the NAACP.
Baltimore County Police claim that Darien had searched for and used artificial resources to mimic students’ voices. He was alleged to have accessed school systems. Darien was also linked to a bogus email address that was used to deliver the recordings, according to the police.
We have previously covered how voice-cloning technology can create realistic conversation after being trained on millions of human voices and therefore tuned to suit a particular voice in a given test. Siwei Lyu, the chairman of the University at Buffalo’s internet forensic lab, spoke with Baltimore Banner reporters in March. According to Lyu, ElevenLabs, a word synthesis company, was responsible for the forged clip of Eiswert speaking. Although its terms of service forbid cloning a message without the person’s consent, ElevenLabs allows users to upload word samples of people for cloning using text-to-talk synthesis.
Since the inspection began, Eiswert has been excluded from the school and has denied making the opinions, claiming that they do not coincide with his beliefs. “I did not make this statement, and these feelings are not what I believe in as both an educator and a man,” Eiswert said in a written statement.
When the voice picture emerged in January, Superintendent Myriam Rogers called the remarks “disturbing” and “highly insulting and inappropriate.” The coalition representing Eiswert, Billy Burke, the only official to formally deny that the voice was AI-generated, according to The Baltimore Banner. He expressed sorrow in the perception that Eiswert’s grief was being shared by the general public and that the director and his family had been threatened and harassed, necessitating police presence at their home.
Not for the first time has AI voice-cloning program caused issues. We’ve previously covered telephone scams where people imitate a loved one’s words in an effort to deceive them into paying them money, and election strategy robocalls that use cloned voices of well-known politicians like Joe Biden. OpenAI revealed the existence of its own voice-cloning systems in March, but the company claimed it was halting it temporarily due to abuse problems.
