Technology media Mac World cited sources on Thursday, stating that Apple has launched two generative AI tools within its corporate firewall as of November 2025—Enchanté, a chatbot, and Enterprise Assistant, a knowledge base application. The goal is to use large models to handle daily queries and content creation for engineers, HR, marketing, and legal teams in one go.
The interface of Enchanté is almost identical to the macOS version of ChatGPT, but all computations are confined to local or Apple's private servers. Internal white lists show that its underlying call range is limited to Apple's self-developed foundation models and third-party models such as Claude and Gemini that have passed security audits. Employee tests showed that by dragging a Swift source code or a Keynote presentation into the window, the system can complete code completion, vulnerability scanning, and text polishing in seconds, and also provide a comparison of executable version differences.
The other tool, Enterprise Assistant, is positioned as a "company-level search engine." Apple has fed dozens of thousands of internal documents, including executive responsibilities, business conduct guidelines, health insurance terms, and device configuration manuals, into a dedicated large language model. Employees can ask questions in natural language and receive answers with paragraph citations, eliminating the need to browse SharePoint or submit IT tickets.
Apple has embedded rating and A/B comparison features in both tools: users can rate each generated result with stars, and they can compare Apple's model outputs side-by-side with those from Claude and Gemini. A source revealed that these feedback data, after being de-identified, directly enter the model training pipeline, used to optimize the upcoming Apple Intelligence features for consumer release.
As of the time of writing, Apple has not responded to requests for comment. Analysts generally believe that amid tight public market competition, Apple is accelerating model iteration through large-scale internal testing, building up "ammunition" for potential new operating system-level AI features expected to be released in 2026.
