Turns out, there’s something even artificial intelligence (AI) can’t do properly—draw a person writing with their left hand. At a time when AI-generated deepfakes are everywhere, Prime Minister Narendra Modi pointed out this strange limitation during his speech at the AI Action Summit in Paris on Tuesday.
PM Modi, who co-chaired the summit with French President Emmanuel Macron, kicked off his speech with an “experiment.”
“If you upload your medical report to an AI app, it can explain in simple language, free of any jargon, what it means for your health,” he said. That’s standard AI functionality—no surprises there.
But then came an interesting observation. “If you ask the same app to draw an image of someone writing with their left hand, the app will most likely draw someone writing with their right hand,” Modi said. The reason? AI is trained on data that mostly features right-handed people.
This got us thinking: Is AI really still making this mistake in 2025, despite all the advancements? We decided to put it to the test.
To verify PM Modi's claim, we conducted a simple experiment using popular AI ChatGPT. When tasked with creating an image of a person writing with their left hand, the AI depicted someone writing with their right hand. Even when asked to generate an image of a person writing with both hands, the AI defaulted to the right hand. This confirmed the Prime Minister's statement that AI struggles with left-handed writing.
So yes, PM Modi was right—AI struggles with left-handed writing. But why?
AI expert Jaspreet Bindra, Co-Founder & CEO of AI&Beyond, explained that this isn’t just a quirk—it’s an example of bias in AI models.
“The Prime Minister picked a brilliant example to show how AI models reflect human biases,” Bindra told us. “Most people in the world write with their right hand, and AI is trained on massive datasets full of such images. That’s why, when asked to generate a left-handed writer, AI still defaults to the right hand.”
And this problem goes beyond handedness. Bias in AI can lead to incorrect assumptions about race, gender, and other attributes.
“For instance, AI systems have mistakenly classified people with darker skin tones as non-human,” Bindra explained. “Similarly, when asked to generate a doctor, AI often defaults to a male doctor because that’s what most of the training data suggests.”
This is exactly why AI ethics is a big deal. While developers work to remove bias, it’s nearly impossible to eliminate it completely—because AI learns from human-created content, which is full of biases.
PM Modi’s speech at the AI Action Summit wasn’t just about a small quirk in AI image generation. It was a reminder that AI is only as fair as the data we feed it. And if we want AI to be truly unbiased, we need to be more careful about how we build and train it.