In an effort to avoid generating false information and ensure accuracy in its AI features, Apple has embedded specific guidelines within its developer beta for macOS 15.1. These guidelines, discovered by a Reddit user, provide a rare glimpse into the backend workings of Apple's AI system, known as Apple Intelligence. The prompts offer detailed instructions for various functionalities, such as the Smart Reply feature in Apple Mail and the Memories feature in Apple Photos.

The primary aim of these prompts is to address the issue of hallucination, preventing the AI from generating inaccurate content. For the Smart Reply feature, the AI is programmed to identify relevant questions from an email and generate concise answers. The prompt instructs the AI to ask the recipient explicit questions from the email, to which the recipient can select answers that will help reduce hallucination in drafting the response. The questions are short, no longer than 8 words, and the answers are kept brief as well, around 2 words.

The Memories feature in Apple Photos, which creates video stories from user photos, follows another set of guidelines to ensure a positive and user-friendly experience. The AI is instructed to generate stories that are free from any controversial, harmful, or negative content. The prompt for this feature involves a conversation between a user requesting a story and a creative writer assistant responding with a story. The generated story should have clear intent, a diverse focus, and a well-defined arc. The photo captions within the story need to adhere to specific guidelines as well.

Apple Intelligence also includes a general directive to avoid hallucination across its AI tools. The Writing Tools feature, for instance, instructs the assistant to provide a concise and natural reply to a user's mail based on a short reply snippet. To make the draft response nicer and complete, a set of questions and their answers are provided, which the assistant incorporates into the reply. The reply is limited to 50 words and must not include any hallucinated or fabricated information.

The rollout of Apple Intelligence is expected to commence officially in iOS 18.1, iPadOS 18.1, and macOS Sequoia 15.1 later this year. Apple users can anticipate the introduction of new features and updates to enhance the AI experience, with continuous improvements expected through to 2025.

By implementing these detailed guidelines, Apple aims to ensure that its AI features deliver accurate, relevant, and user-friendly content, while also avoiding the generation of false information. As Apple Intelligence becomes more integrated into users' daily lives, these measures will play a crucial role in maintaining the integrity and trustworthiness of the AI system.