Grok Misidentifies Gaza Photo as Yemen Image From 2018
Grok, an AI chatbot integrated into X, falsely identified a photograph of a malnourished child from Gaza taken in August 2025 as a 2018 Yemen image, misattributing it to a different child. The chatbot repeated the error even after acknowledging it was wrong, raising concerns about AI reliability in image verification.
What happened
When French MP Aymeric Caron published a photo of a malnourished child on X in early August 2025, users asked Grok to verify its origin. Grok stated the photo was taken in October 2018 in Yemen and showed a 7-year-old girl named Amal Hussain.
The photo was actually taken on August 2, 2025 in Gaza by photojournalist Omar al-Qattaa of Agence France-Presse, showing 9-year-old Mariam Dawwas in her mother's arms. When questioned about the error, Grok initially claimed it did not spread false news and relied on verified sources, then admitted the mistake but repeated it the following day when asked by other X users.
Reported impact
Grok's misidentification was widely shared, leading to accusations that Caron was spreading disinformation about Gaza. The error illustrated limitations in AI image verification, with researchers noting that language models function as 'black boxes' with biases from training data and designer instructions.
Index analysis
Grok's errors reflected underlying problems with AI tools in verifying images. Expert Louis de Diesbach explained that when asked about image origins, language models may generate plausible but incorrect answers, such as suggesting an image could come from any location experiencing famine.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.