Facebook's automated hate speech detection system mistakenly removed a post containing excerpts from the Declaration of Independence that included the phrase 'Indian Savages', which was flagged as hate speech before being restored after human review.
Facebook's automated content moderation system removed a post by The Vindicator, a small community newspaper in Liberty County, Texas, that contained excerpts from the Declaration of Independence posted in preparation for the Fourth of July. The post contained the phrase 'merciless Indian Savages' from the original historical document, which Facebook's automated systems flagged as hate speech and removed. The incident occurred in early July 2018 when the newspaper was posting daily excerpts of the Declaration of Independence. After The Vindicator's editors published a story about the removal on July 3 and notified Facebook, the company restored the post and apologized. Facebook acknowledged that the post was 'removed by mistake' and explained that their automated systems likely detected the phrase 'Indian Savages' and triggered the removal. A Facebook spokesperson stated that they 'process millions of reports each week, and sometimes we get things wrong.' The incident highlights ongoing challenges with Facebook's content moderation system, which uses a combination of humans and automation to review posts, and demonstrates the difficulty of distinguishing between legitimate historical content and actual hate speech.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that fail to perform reliably or effectively under varying conditions, exposing them to errors and failures that can have significant consequences, especially in critical applications or areas that require moral reasoning.
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