A 66-year-old woman from Dunfermline, Scotland, Louise Littlejohn, was left shocked after Apple’s AI-powered dictation software turned a routine voicemail into an offensive message. The voicemail, originally from Lookers Land Rover garage in Motherwell, was meant to be a simple invitation to an upcoming event. However, when transcribed by Apple’s AI, it included inappropriate language, referring to her with an insult and making an unsolicited remark about her personal life.
According to the BBC, the AI mistakenly included offensive language, calling her a "piece of s**t" and making an inappropriate remark about her personal life. Littlejohn, shocked by the error, described the experience as "obviously quite inappropriate." The incident has raised concerns over the accuracy and reliability of AI-driven transcription services, especially when dealing with sensitive communications.
Littlejohn initially thought the message was a scam. However, after recognizing the caller’s area code, she realized it was from the garage that had previously sold her a car. They were simply following up, but the AI-transcribed voicemail turned the routine call into an offensive message.
She was quick to clarify that the garage itself was not to blame, pointing the finger at AI’s flawed voice recognition instead. “The garage is trying to sell cars, and instead of that, they are leaving insulting messages without even being aware of it,” she said.
Experts believe that multiple factors could have contributed to the bizarre mistake. While some suggest the caller’s Scottish accent may have played a role, others argue that background noise, a scripted call, or even AI’s inherent limitations in understanding speech could have caused the blunder.
Peter Bell, a professor of speech technology at the University of Edinburgh, explained that speech recognition systems are not perfect and can struggle with accents, background interference, and varying speech patterns. “All of those factors contribute to the system doing badly,” he told the Daily Mail.
BBC tech analysts believe the AI may have misheard the word “sixth” (referring to the sixth of March, the event date) as something inappropriate, leading to the mistake. It was a classic case of AI misinterpretation with an unexpected twist.
Though Littlejohn was initially shocked, she later found the situation amusing. “At first, I couldn’t believe it, but then I just had to laugh,” she said. While this mistake was more funny than harmful, experts say it points to a bigger issue. AI transcription is not perfect and can sometimes create inappropriate messages.“The main concern is why the AI chooses such words.
A system used by the public should have protections to stop this,” said Peter Bell.
Apple explained that these mistakes happen because some words sound similar. Sometimes, the software shows the wrong word before fixing it. Clearly, AI still has a lot to learn about human speech. Until then, it's a good idea to check transcriptions before sending important messages.
You might also be interested in - Flight forced to return after passenger claims evil spirits followed him, attacks crew