Powered by
MOMENTUM
MEDIA
The darker side of artificial intelligence has been exposed in a new study, which reveals that AI tools designed to generate images and videos are distorting the image of the legal profession, with women and racial minorities largely absent.
Artificial intelligence tools designed to generate images and videos are failing to capture the true diversity of the legal profession, with women and racial minorities significantly underrepresented, according to a new analysis.
A recent study by Kapwing, sampling AI-generated videos from platforms including Google’s Veo 3, OpenAI’s Sora 2, Kling, and Hailuo Minimax, revealed how these cutting-edge tools depict lawyers and judges.
The analysis revealed a stark disparity between AI-generated depictions and the real-world demographics of the legal profession, finding that only 21.6 per cent of lawyers were depicted as women, despite women making up 53 per cent of Australia’s legal workforce.
In some cases, Kapwing also found that some of the most popular AI video tools “failed to depict any women” as lawyers at all.
The analysis showed that female judges were also misrepresented, with several AI video tools heavily favouring male depictions and underrepresenting women in the judiciary.
Liam Curtis, content strategist at Kapwing, explained that the disparity stems from the AI’s training data, which reflects historical biases in which men have long dominated media and stock imagery, meaning these tools are essentially replicating existing societal inequalities.
“The disparity likely reflects patterns in the data these models were trained on. AI models learn from large volumes of internet content, and historically, media and stock imagery have overrepresented men in certain professions,” he said.
“It’s likely an extension of biases that already existed in the underlying data.”
However, the issue goes beyond gender, with Kapwing finding that AI-generated videos also perpetuate the overrepresentation of white individuals among lawyers and judges.
The study revealed a stark racial bias, showing that nearly 82 per cent of AI-generated lawyers were depicted as white, with judges similarly portrayed as “overwhelmingly white” in multiple models.
Curtis warned that as AI-generated visuals become a new layer of the media landscape, depicting lawyers and judges as predominantly male and white could reinforce outdated stereotypes and shape public perceptions of who belongs in the legal profession.
“AI-generated visuals are becoming another layer in the media ecosystem – alongside advertising, film, television and stock imagery – where representation shapes perception,” he said.
“Without intervention, these systems may replicate and scale the same biases that have historically appeared in other visual media.”
To prevent AI from simply perpetuating historical biases, Curtis emphasised that human oversight is critical, with AI treated as a creative assistant rather than a final decision-maker, and outputs carefully reviewed and refined to reflect real-world diversity.
“I think it reinforces the importance of human oversight. AI-generated content should be reviewed and iterated on, not published automatically,” she said.
“Organisations should use tools that allow for previewing, refining and adjusting outputs to better reflect real-world diversity. Responsible use means treating AI as a creative assistant, not a final decision-maker.”
Full report is available here.
