🧭 Claude Joins Apple's Foundation Models Framework at WWDC 2026

At WWDC 2026 (June 9–13, Cupertino), Apple and Anthropic jointly announced that Claude is now part of Apple's Foundation Models framework — the on-device AI infrastructure introduced with Apple Intelligence. Starting with iOS 27, iPadOS 27, macOS 27, visionOS 27, and watchOS 27 (available this autumn), developers can use a new Swift package to route tasks to Claude when they exceed what the on-device model can handle — including multi-step reasoning, code generation, web search, and code execution. Streaming responses are native to SwiftUI, so real-time Claude output in iOS and macOS apps requires no custom transport layer.

How the framework integration works

• Automatic delegation: The Foundation Models framework decides whether a prompt stays on-device or routes to Claude based on task complexity and connectivity. Developers can also set explicit routing policies via the InferenceTarget API.
• Typed Swift outputs: Claude responses are decoded directly into Swift structs conforming to Codable, so the app receives structured data without manual JSON parsing.
• Operator control: Anthropic's operator system-prompt layer is preserved — app developers can pass a systemPrompt to configure Claude's persona, constraints, and response style.
• Private relay: Traffic between device and Claude API is routed through Apple's privacy relay infrastructure, so Anthropic does not see device identifiers or Apple account information.

import FoundationModels
import Anthropic  // new Apple-distributed Swift package

// Route a complex reasoning task to Claude automatically
let session = LanguageModelSession(
    model: .automatic,           // Foundation Models picks on-device or Claude
    instructions: "You are a helpful coding assistant."
)

let response = try await session.respond(to: "Refactor this Swift function to use async/await: \(code)")
print(response.content)

// Force Claude explicitly for tasks that need internet reasoning
let claudeSession = LanguageModelSession(
    model: .claude,
    instructions: "You are a senior Swift engineer."
)
for try await chunk in claudeSession.streamResponse(to: "Explain this crash log: \(log)") {
    textView.append(chunk)  // native SwiftUI streaming
}

🧭 LG CNS Signs Group-Wide Claude Enterprise Agreement — Asia's Largest Single-Org Deployment

LG CNS — the IT-services arm of South Korea's LG Group, with roughly 7,000 employees and annual revenue above $2 billion — has signed a group-wide Claude Enterprise agreement with Anthropic, covering all LG affiliates from LG Electronics and LG Energy Solution to LG Display and LG Uplus. The deal, reported in the Korea Herald and Seoul Economic Daily, makes it one of the largest single-organisation Claude deployments in Asia to date. The agreement was timed to coincide with Code with Claude Tokyo (June 10–11), signalling Anthropic's commitment to the Asia-Pacific enterprise market.

Planned use cases across LG affiliates

• Agentic software development: Claude Code running in LG CNS-managed sandboxes to accelerate internal software projects across the group — aligned with LG CNS's existing DevSecOps platform.
• AI agent creation for non-technical teams: Business units at LG affiliates will use Claude to build department-specific agents without needing to write code, using Claude's natural-language workflow tools.
• Long-document processing: Legal, compliance, and procurement contracts (often 200–500 pages) processed through Claude's 1M-token context window for summarisation, risk flagging, and cross-document comparison.
• Korean-language optimisation: Anthropic confirmed the enterprise agreement includes an SLA commitment on Korean-language quality, including formal register (존댓말 / jondaemal) fidelity in business correspondence drafts.

🧭 Code with Claude Tokyo: Full Schedule Released — What to Watch June 10–11

With Code with Claude Tokyo opening at 09:00 JST tomorrow at the Tokyo International Forum, Anthropic has published the full two-day session schedule. The conference is Anthropic's first major developer event in Asia, and the lineup reflects the region's distinct priorities: strong enterprise integration tracks, Japanese-language model optimisation, and hands-on workshops targeted at iOS/Swift and Android/Kotlin developers now that Claude is embedded in Apple's Foundation Models framework. Remote attendance is free; all keynotes and selected sessions will be streamed live on anthropic.com.

Highlighted sessions (selected)

• Opening keynote — Dario Amodei: Live demos of Claude Code running agentic workflows at enterprise scale, followed by a Q&A segment. June 10, 09:15 JST.
• "Scaling Agentic Workflows at Enterprise" panel: Practitioners from LG CNS, Rakuten, and Softbank share what's working and what isn't in large-scale Claude deployments. June 10, 11:00 JST.
• "Claude for Japanese Business Writing" session: Deep dive on formal register, keigo accuracy, and document-formatting norms — with benchmark comparisons between Claude Opus 4.8 and competing models on Japanese business prose. June 10, 14:00 JST.
• MCP Server Workshop: Build and deploy a private MCP server — accessible only to your organisation's Claude agents — in 90 minutes. Limited to 80 in-person attendees; materials open-sourced on GitHub. June 10, 15:30 JST.
• "Mobile AI with Claude" (iOS/Android): Hands-on session covering the Foundation Models Swift package integration and the new Android AI Core bindings. June 11, 10:00 JST.
• Closing keynote — research preview: Anthropic researchers will share a preview of upcoming model capabilities targeted at Asia-Pacific use cases. Details embargoed until broadcast. June 11, 16:00 JST.