Apple’s iOS Keyboard Blocks Autocomplete for Sensitive and Vulgar Terms

Apple maintains a secretive multilingual gray list containing tens of thousands of words and phrases that its iOS keyboard will not autocomplete or allow via swipe-typing, according to independent researcher Derik Kauffman. This list is designed to prevent certain offensive, vulgar, or politically sensitive terms from being suggested, although users can still manually type these words.

Kauffman uncovered this feature after observing that some phrases, like “bad boy” or “white man,” caused swipe-typing to hesitate, while similar phrases such as “bad girl” or “black man” did not. His investigation led him to discover a vast database of restricted words embedded deep within Apple’s operating system. Estimates suggest the list includes around 100,000 entries, excluding Chinese characters, spanning multiple languages.

The words flagged are mostly those related to obscenity, drug use, sex, or violence. However, the list also contains politically charged terms such as “dictator Trump,” “sleepy Joe,” “right wing conspiracy,” “Tibetan independence,” and “United Ireland.” According to Kauffman, similar lists exist for other major languages, with French users unable to swipe-type phrases like “immigration too high,” and Hebrew speakers restricted from swiping “occupied territories.”

Attempts to swipe-type these flagged phrases often result in incorrect or altered suggestions. For example, swiping “United Ireland” might produce “United Iceland,” while “Tibetan independence” becomes “Tibetan independences.” Additionally, “white man” converts to “white mean,” whereas “red man” and “blue man” are accepted. The phrase “sleepy Joe” is changed to “sleepy hope,” and “read gay smut” is blocked while “read smut” works normally.

Apple has not officially commented on the exact contents or the rationale behind this list but stated that the keyboard is designed to avoid suggesting inappropriate or offensive content. This selective filtering represents a cautious approach to moderating language suggestions while preserving user freedom to type what they wish.

Matt Suiche, a cybersecurity engineer at the company Tolmo, independently confirmed Kauffman’s findings by retrieving the same gray list and verifying it against Apple’s internal data. Suiche noted that this extensive filtering has likely influenced how users communicate on Apple devices over many years, pointing to a subtle form of content moderation embedded within everyday typing experiences.

While Apple’s intention to reduce the chance of users accidentally sending offensive or inflammatory messages is clear, Kauffman pointed out inconsistencies in the implementation. Some phrases are blocked while very similar ones are allowed, which indicates a somewhat imprecise filtering process. He described it as “a very light-touch form of censorship” that can still have a significant impact given Apple’s vast global user base.

This discovery arrives amid ongoing discussions about transparency and control over content moderation by major technology companies. Social media users have long noticed that certain phrases cannot be swipe-typed or suggested, but the full scale of Apple’s gray list has not been publicly documented until now. Recently, a Norwegian systems administrator also found a subset of blocked words, including terms like “al-Qaeda,” “Auschwitz,” and “Hooters.”

Apple has faced criticism in the past over predictive text features. For instance, Google’s autocomplete has frequently been scrutinized for suggesting controversial or offensive search terms. In 2023, Apple drew media attention when iPhone users noticed that dictating the word “racist” briefly displayed “Trump” on their screens. Apple attributed this to a software error and corrected it promptly.

David Greene, senior counsel at the Electronic Frontier Foundation, highlighted the tension between protecting users from harmful content and the risks of hidden censorship. While Greene acknowledged that filtering offensive suggestions could be beneficial, he argued that users should be fully informed and have options to customize these settings. He emphasized that such restrictions should not remain undisclosed.

As millions of people worldwide rely on Apple’s keyboard for daily communication, the way the company manages suggested text has a subtle but meaningful influence on language use and expression. Greater openness about these content moderation practices and offering users more control could help balance safety with freedom of expression.