AI-generated 'poverty porn' fake images being used by aid agencies (www.theguardian.com)

AI-generated images depicting extreme poverty, children and survivors of sexual violence are proliferating on stock-photo platforms and being used in campaigns by some NGOs, prompting alarm from global health professionals who call the phenomenon “poverty porn 2.0.” Researchers have documented dozens of synthetic images that recreate a familiar “visual grammar” of poverty — malnourished children, cracked earth, racialized scenes — sometimes licensed through major sites such as Adobe Stock and Freepik. High-profile missteps include a 2023 Plan International campaign that used AI images to avoid showing real children, and a UN video that was taken down after it mixed near-real AI-generated reenactments of sexual violence with real footage. The trend matters technically and ethically. Generative models reproduce and exaggerate societal biases, so widespread synthetic stereotypes risk normalizing demeaning portrayals, retraumatizing subjects, and eroding consent norms; they’re also cheap to produce, incentivizing use amid NGO budget cuts. Critically, these biased images can leak back into the web and training datasets, creating a feedback loop that amplifies prejudice in future models and degrades information integrity. The debate is shifting from real-world “poverty porn” to whether and how the sector should permit synthetic portrayals at all, with some organizations now adopting guidance to avoid AI depictions of individual children.