Facebook AI Mistranslation of Arabic Post Leads to Wrongful Arrest in Israel (2017)

Incident Summary

In October 2017, Israeli police detained a Palestinian construction worker after Facebook’s automatic translation feature mistranslated his Arabic-language post from “Good morning” to “Attack them” in Hebrew and “Hurt them” in English.^[1][2] The man had posted a photograph of himself next to a bulldozer at his workplace with the caption containing the Arabic word “yisbahhum,” a colloquial form of “good morning.”^[1]

Israeli police, who monitor social media posts, interpreted the mistranslated text as a threat and detained the man for several hours before the error was identified.^[2][3] Facebook acknowledged the translation error, attributing it to limitations in its AI-powered translation system’s handling of Arabic dialect variations.^[1]

Key Facts

• Affected individual: A Palestinian construction worker in Beitar Illit, Israel
• Original post: Arabic-language caption meaning “Good morning” alongside a workplace photograph
• Mistranslation: AI translated Arabic text to “Attack them” (Hebrew) and “Hurt them” (English)
• Consequence: Man detained by Israeli police for several hours
• Resolution: Translation error identified; man released
• Platform response: Facebook acknowledged the error and stated it was working to improve translation accuracy

Threat Patterns Involved

Primary: Misinformation and Hallucinated Content — The AI translation system generated a factually incorrect and dangerously misleading output, transforming an innocuous greeting into an apparent threat of violence.

Secondary: Proxy Discrimination — The mistranslation disproportionately affected an Arabic speaker in a geopolitically sensitive context where social media posts are actively monitored by law enforcement, illustrating how AI errors can compound existing power asymmetries.

Significance

1. Real-world consequences of AI translation errors. The incident demonstrated that AI translation failures can result in immediate, tangible harm — including wrongful detention — when automated outputs are treated as reliable by law enforcement.
2. AI errors in high-stakes surveillance contexts. When AI translation systems operate in environments where social media is actively monitored by security forces, even routine errors can escalate rapidly to deprivation of liberty.
3. Arabic language processing challenges. The incident highlighted the limitations of AI translation systems in handling dialectal Arabic, where colloquial forms can differ significantly from Modern Standard Arabic and may be misinterpreted by models trained on inadequate data.
4. Accountability gap. Neither Facebook nor the law enforcement agencies that relied on the mistranslation faced formal consequences, raising questions about accountability when AI errors lead to wrongful detention.