Facebook and Twitter's AI content moderation systems wrongly blocked millions of pro-Palestinian posts and accounts during the 2021 Israeli-Palestinian conflict, with algorithms mistakenly identifying legitimate content as spam or hate speech.
During the May 2021 Israeli-Palestinian conflict, Facebook and Twitter's AI content moderation systems caused widespread censorship of pro-Palestinian content. Twitter's AI mistakenly identified rapid tweeting during confrontations as spam, resulting in hundreds of accounts being temporarily locked. Facebook's hate speech detection software misidentified the #AlAqsa hashtag as associated with a terrorist group, and a software bug temporarily blocked video-sharing on Instagram. The companies said problems were quickly resolved, but activists reported many posts remained censored. Facebook employees filed at least 80 tickets reporting 'false positives' with automation systems, noting AI was mislabeling protest images as 'harassment or bullying.' The incident was designated 'severity 1' by Facebook. According to 7amleh advocacy group, 170 Instagram posts and 5 Twitter posts remained offline even after companies claimed fixes were implemented. The incident highlighted broader issues with algorithmic bias affecting marginalized groups, with similar problems documented against Palestinian content since 2016.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Unequal treatment of individuals or groups by AI, often based on race, gender, or other sensitive characteristics, resulting in unfair outcomes and unfair representation of those groups.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed