Apple Restricts Over 100,000 Everyday Terms on the iPhone Keyboard

2026-11-07 companies

Cupertino, Saturday, 7 November 2026.
Researchers uncovered a hidden database blocking more than 100,000 political, vulgar, and sensitive terms from iPhone text assistance, spotlighting device-level content moderation and quiet corporate influence across global digital communication.

Discovery of Apple’s Hidden Grey List

An independent investigation published on November 5, 2026, revealed that Apple Inc. (NASDAQ: AAPL) maintains an internal “grey list” containing over 100,000 restricted words and phrases that its iOS predictive text engine suppresses from auto-completing or swipe-typing functions [1][2][4]. The list, discovered by researcher Derik Kauffman in September 2026, was unearthed after he experienced unusual inconsistencies while using the iPhone’s swipe-typing feature [1][2][5]. Kauffman described the filtering as a ‘light-touch form of censorship,’ noting that while individual restrictions may seem minor, their cumulative impact at Apple’s scale is significant [1][2][5].

Scope and Categories of Restricted Terms

The grey list spans multiple languages and includes categories such as vulgarities, sexual content, drug references, violence, political slogans, historical events, and public figures [1][2][3][4]. Specific examples of restricted terms include political phrases like ‘Dictator Trump,’ ‘Sleepy Joe,’ ‘Tibetan independence,’ ‘United Ireland,’ ‘Modi braggart,’ and ‘Rahul is Pappu’ [1][2][3][4]. Vulgar terms such as ‘male chauvinist pig’ and drug-related phrases like ‘read gay smut’ are also blocked [1][2][4]. Additionally, seemingly mundane phrases like ‘iMessage isn’t working’ and ‘Santa isn’t real’ have been flagged, demonstrating the breadth of Apple’s filtering criteria [1][3][4]. Language-specific restrictions are evident, such as blocking ‘the occupied territories’ in Hebrew and ‘the Hamas movement’ in Arabic [1][2][4].

Technical Evolution and User Impact

Apple’s predictive text system has evolved significantly over the years, transitioning from Unix word-lookup tables to statistical n-grams and, in 2025, to a transformer-based autocorrect system akin to a large language model (LLM) [1][GPT]. This latest system, launched in 2025, has been linked to unusual spelling suggestions and autocorrect alterations reported by users [1][2][GPT]. Despite these restrictions, users can still manually type any of the blocked terms, as the grey list only affects predictive text and swipe-typing features [1][2][3][4]. However, the filtering has led to inconsistencies, such as allowing ‘read smut’ while blocking ‘read gay smut,’ and permitting ‘bad girl’ while restricting ‘bad boy’ [2][5].

Broader Implications and Corporate Accountability

The revelation of Apple’s grey list underscores the quiet yet pervasive influence of corporate algorithmic moderation on digital communication [1][2][4][5]. David Greene, senior counsel at the Electronic Frontier Foundation, emphasized that if companies like Apple choose to implement such filters, users should be fully aware of them and have the ability to customize or opt out [2]. The grey list operates across all major languages, reflecting cultural and regional sensitivities, and its existence raises questions about transparency, user agency, and the ethical responsibilities of tech giants in shaping public discourse [1][2][3][4]. As of November 7, 2026, Apple has not publicly confirmed the size or specific contents of the grey list, stating only that its keyboard is designed to avoid inappropriate suggestions [1][2][3][4].

Sources

Algorithmic Moderation Device Governance