Apple Accidentally Leaves Some of Its AI Prompts Out in the Open: ‘Do Not Hallucinate’

Like any API in large language models (LLM), Apple’s AI system, known as Apple Intelligence, requires prompts before users can query it. MacRumors reports that a Reddit user has discovered these prompts, offering a peek into how Apple’s AI works behind the scenes.

Why this matters. These guidelines demonstrate how Apple handles queries to its AI model with regard to privacy, accuracy, and ethics.

Some context. A Reddit user found files in the beta version of macOS Sequoia 15.1 that contain pre-prompt instructions for some Apple Intelligence features. The company puts these guidelines in place for processing user requests.

Key points. Some of these guidelines include:

• Identity and core values: “You are an AI assistant created by Apple. You are designed to be helpful, harmless, and honest.”
• Privacy: “Respect personal privacy. Do not ask for or share personal information.”
• Accuracy: “If you’re unsure about something, say so. Don’t make up information or pretend to know things you don’t.”
• Generating email responses: The assistant is instructed to identify relevant questions from an email and generate concise responses, avoiding “hallucinations” (generation of false information presented as authentic by an LLM).
• Creation of memory videos: From the user’s photos, the assistant should generate positive stories free of controversial or harmful content.

Perhaps the most striking instructions relate to hallucinations. Specifically, to avoid them. This is the full prompt:

“You are an assistant which helps the user respond to their mails. Given a mail, a draft response is initially provided based on a short reply snippet. In order to make the draft response nicer and complete, a set of question and its answer are provided. Please write a concise and natural reply by modifying the draft response to incorporate the given questions and their answers. Please limit the reply within 50 words. Do not hallucinate. Do not make up factual information.”

It’s also interesting to read the full prompt for the smart answer feature:

“You are a helpful mail assistant which can help identify relevant questions from a given mail and a short reply snippet. Given a mail and the reply snippet, ask relevant questions which are explicitly asked in the mail. The answer to those questions will be selected by the recipient which will help reduce hallucination in drafting the response. Please output top questions along with set of possible answers/options for each of those questions. Do not ask questions which are answered by the reply snippet. The questions should be short, no more than 8 words. The answers should be short as well, around 2 words. Present your output in a json format with a list of dictionaries containing question and answers as the keys. If no question is asked in the mail, then output an empty list. Only output valid json and nothing else.”

Reading between the lines. The pre-prompts demonstrate Apple’s careful approach, which isn’t far from the instructions that any user familiar with these tools might include on their own. The key takeaways are:

• Emphasis on user privacy and security.
• Concern for avoiding the generation of false or misleading information.
• Attempt to maintain a positive and non-controversial tone in the content.

What’s next. The first Apple Intelligence features are just starting to hit the betas of the newly released operating systems, but will be available to all U.S. users with iOS 18.1, iPadOS 18.1, and macOS Sequoia 15.1 in the fall, likely in October.

The company will introduce Apple Intelligence to other regions and features as early as 2025. While they wait, users can take a look at the hidden script that Apple has prepared to define its own AI system.

This article was written by Javier Lacort and originally published in Spanish on Xataka.

Image | Apple

Related | Google’s New Monopoly Status Has Unexpected Collateral Damage: Apple