A recent discovery hints that Apple may be using Anthropic’s Claude internally. The company’s latest Apple Support app update reportedly shipped with internal CLAUDE.md files.<\/p>\n
Researcher and analyst Aaron Perris flagged the files on X (formerly Twitter).<\/p>\n
Apple’s AI Workflow Leaked?<\/p>\n
CLAUDE.md is a feature of Claude Code, Anthropic’s AI coding assistant. Developers drop these markdown files into their project directories, and Claude Code reads them at the start of every session.<\/p>\n
They’re used to set coding standards, define architecture decisions, specify preferred libraries, and define review checklists.<\/p>\n
They typically live in source repos and are not meant to ship inside production apps,\u201d International Cyber Digest wrote. \u201cOh, and this also confirms Apple is using Claude internally.\u201d<\/p>\n
In the shared screenshots, the first file describes a chat module that combines Juno AI with live human agents. It outlines three participant roles for message routing: client, agent, and assistant.<\/p>\n
Notably, the architecture relies on conditional compilation flags such as JUNO_ENABLED and DEV_BUILD. The file also references an internal Apple bug-tracker entry, indicating it originates from a live codebase.<\/p>\n
A second file documents SAComponents, a shared UI library that supports multiple Apple platforms, including visionOS. It separates UI from business logic and supports both SwiftUI and UIKit.<\/p>\n
Claude files surfacing in a public build may signal a packaging error during release. It could also hint that Apple’s engineers may be incorporating Claude into how they build software internally.<\/p>\n
Apple’s Public Ties to Anthropic Run Deep<\/p>\n
Apple previously confirmed support for Claude Sonnet 4 in Xcode 26 in September 2025. Then, in February 2026, the company expanded this with Xcode 26.3. The update added native integration with the Claude Agent SDK.<\/p>\n
The company is widely recognized for its tight control over hardware, software, and security. That self-contained approach has long shaped Apple’s reputation as one of the most security-focused product makers in the industry.<\/p>\n
The shipped CLAUDE.md files now indicate that Anthropic’s tools may be sitting alongside those resources, raising fresh questions about how Apple balances third-party AI use with its tightly controlled internal stack.<\/p>\n
Subscribe to our YouTube channel to watch leaders and journalists provide expert insights<\/p>\n