"AI is biased": at the Cyprus Forum, the gender equality commissioner puts the training data on trial

Artificial intelligence is already reproducing human biases and gender stereotypes, which makes who sits in the room designing it a technical question rather than a social courtesy. That was the argument Gender Equality Commissioner Josie Christodoulou brought to the Cyprus Forum in Nicosia, in a fireside discussion titled "AI and Equal Participation: Who Shapes the Digital State?"

Her claim was blunt — "AI is biased" — and she located the cause in two places: the data the systems are trained on, and the people who build them. The illustration she used was deliberately simple. Give a model a thousand books in which the doctor is a man and the nurse is a woman every time, and the model will hand that back to you as fact.

The problem compounds where women are thinly represented in technology and innovation, because the people developing a system inevitably decide which experiences it accounts for. Christodoulou also pointed to a retention pattern rather than a recruitment one: women do enter tech, then leave when they start families — a cultural outcome, she argued, not a limit on what they can do.

Her proposed remedies ran across the pipeline: encourage girls into STEM through education, design technology that deliberately serves women, and put women into the decision-making that shapes it. She closed on child safety, flagging AI tools that enable deepfake abuse and sexualised imagery of minors, and arguing for digital literacy paired with human oversight.

The panel was chaired by Andreas Demetriou, a founding member of AI³, and co-chaired by Rodoula Papalambrianou of the Hope For Children CRC Policy Center.

Source