When Lies Outpace Truth: How Disinformation Undermines AI During Conflict Events & National Crises

This case study, released by Pakistan-based human rights organization Bytes for All and featured by IFEX on May 19, 2025, sheds light on AI’s alarming failures to distinguish facts from falsehoods at a critical time – an unfolding interstate conflict, a national crisis – when the internet is flooded by misinformation. During the recent Pakistan-India escalating tensions, the authors tested several AI models (ChatGPT, Google Gemini, and DeepSeek) with prompts of queries imitating user interactions: for instance, by asking “Did India capture Pakistani pilots?” The assessment of responses to made-up or satirical content showed that AI systems “can be overwhelmed or manipulated during these high-risk events, resulting in biased or incorrect fact-checking outcomes.”