Extensive Blocklist Uncovered in iOS
An examination of Apple iOS has revealed an internal filter containing more than 100,000 words and phrases blocked from predictive text suggestions and swipe-typing features. The discovery, made by researcher Derik Kauffman and reported by Reuters, demonstrates how Apple limits automated keyboard suggestions across every major language supported by the operating system.
According to Kauffman’s findings, the restricted database does not prohibit users from manually entering the targeted terms. iPhone owners can still type any phrase letter by letter without restriction. However, the system’s software will not surface the entries in the predictive text banner, and the iOS swipe-to-type engine will frequently redirect inputs to alternative words. For example, attempting to swipe the phrase “left wing conspiracy” may result in the keyboard automatically altering the text to “left wrong conspiracy.”
Political Phrases, Slurs, and Benign Entries
The vast majority of the cataloged entries encompass explicit content, slurs, offensive terminology, and references to drugs or sexual acts. Alongside typical content moderation triggers, the database also includes a wide array of politically charged expressions and references to prominent figures.
Among the political terms withheld from autocomplete and swipe typing are phrases such as “dictator Trump,” “sleepy Joe,” “illegal immigrant,” “covid was a hoax,” and “left wing conspiracy.” The filter also restricts seemingly non-political or harmless statements, including “Santa isn’t real” and “iMessage isn’t working.”
Kauffman also highlighted several apparent inconsistencies across the compiled data. In some instances, a specific gendered phrasing is suppressed while its direct counterpart remains uninhibited. For example, the phrase “bad girl” appears on the blocked list, whereas “bad boy” is absent and remains available for standard prediction and swipe entry.
Technical Extraction via AI Coding Tools
To uncover the suppression directory from within iOS, Kauffman utilized artificial intelligence developer tools, including OpenAI’s Codex and Anthropic’s Claude Code. Kauffman described locating the data deep within the operating system’s architecture.
On his website, where the full lists for multiple languages have been published, Kauffman noted that the banned terms were compiled and stored in the form of a hash table. By reverse-engineering and resolving the hash structure, he was able to convert the internal records back into readable plain text. The authenticity of the dataset was subsequently supported when another independent researcher confirmed to Reuters that they had obtained an identical list from the software.
Apple Defends Keyboard Design
Addressing the findings, Apple stated to Reuters that the iOS keyboard system is specifically engineered to prevent the software from generating potentially offensive, harmful, or inappropriate suggestions unprompted. The company stressed that the underlying feature does not restrict communication or prevent people from expressing themselves, pointing out that device owners can manually type any phrase they choose or override the software’s automated suggestions.
While Apple maintains that users preserve full freedom to write whatever they wish, the discovery provides a comprehensive view into the boundaries and specific language parameters established inside one of the world’s most widely deployed mobile keyboards.
