Apple's notification summaries feature generates inaccurate text, sometimes inventing words that don't exist. The AI system, part of Apple Intelligence, summarizes lengthy notifications into brief previews. Parents relying on these summaries to quickly scan messages about their children may receive garbled or false information.
The problem stems from how large language models work. These systems predict the next word based on patterns in training data, not by retrieving exact text. When asked to condense information, the AI can confuse details or fabricate plausible-sounding words that fit grammatically but carry no real meaning.
Apple released this feature in iOS 18.1 as part of a broader push into generative AI. The company claims summaries help users process information faster. Early users have documented numerous errors, from subtle word swaps to completely invented terms.
For parents, this matters. School alerts, medical appointment reminders, and messages from childcare providers flowing through Apple devices may arrive distorted. A notification about pickup times or health concerns could contain inaccurate details that parents miss on first read.
Apple has not announced plans to disable the feature or fix the hallucination problem. Parents using iPhones should verify critical information by opening full notifications rather than trusting summaries alone.