The Aftenposten Modi Cartoon: The Image Will Fade. The AI Data Won't.
- Renuka Naj
- Jun 1
- 2 min read
When Norway's largest newspaper published a cartoon of India's Prime Minister as a snake charmer, the world debated racism and press freedom. The more consequential question, what happens when that image enters AI training data, went unasked.
Aftenposten drew India's Prime Minister as a snake charmer, flute in hand, with a fuel pipeline rising from the basket like a cobra. The accompanying piece, by journalist Frank Rossavik, was intended as commentary on India's energy diplomacy. The caricature went viral. Politicians, diaspora communities, and commentators condemned it. Aftenposten expressed regret, describing the reaction as a misreading of the image's intent.
The timing made it worse. Days earlier, a Norwegian journalist had asked Prime Minister Modi why he would not take questions from what she called the freest press in the world. It was a fair question. Then her own country's press answered it for him in a single drawing, and not in the way she meant.
The snake charmer is an old colonial image, the kind once used to paint India as exotic, mystical, and not quite modern. It does not fit in 2026. India has landed near the Moon's south pole, built digital systems other governments now study, and earned a real voice in how AI gets governed.
Aftenposten had good intentions and poor instincts. What a newsroom believes it is saying and what the world receives are two different things. Editorial judgment lives in that gap.
Most coverage stopped at the racism, and rightly so. But the more durable problem is this: stereotypes no longer disappear with the news cycle. AI systems are trained on the media record, and this cartoon is now part of that record. When a model encounters repeated associations between India and colonial imagery, those associations compound over time, becoming stronger features of the data from which future outputs are drawn. A model can be fine-tuned to flag inappropriate imagery. Reversing the underlying association, formed across years of training, requires far greater effort.
That is why diverse, globally representative training data is the foundation on which reliable AI is built. For multilateral institutions and governments, the representation of the Global South in the media has become an issue of governance. AI systems trained on archives that consistently portray emerging economies through outdated Western lenses carry those distortions forward, into translation tools, search results, and digital assistants used by billions of people. That harm accumulates over time.
I have spent more than 20 years in journalism, diplomacy, and communications, living in places like Bangladesh, South Sudan, Iraq, and Ethiopia. Working across multiple knowledge systems builds a specific kind of judgment: the ability to recognise when an assumption that appears neutral inside one framework carries entirely different weight in another. That cross-cultural fluency is a technical requirement now for building AI that works accurately and equitably across the full range of human experience.
The cartoon will be forgotten. But archive will remain. The AI systems trained on today's media will carry our blind spots forward just as they carry our knowledge. The judgment we exercise today is the intelligence we leave behind for future generations.
Comments