A new Apple study reveals a significant flaw in large language models like ChatGPT: their accuracy plummets when faced with irrelevant information. This raises questions about AI's true comprehension abilities and the need for caution in relying on these systems for complex tasks.