Anthropic Opens Seoul Office, Lands NAVER, Samsung SDS, and LG CNS
Anthropic formally opened its Seoul office — its third in Asia-Pacific after Tokyo and Bengaluru — led by KiYoung Choi, a three-decade veteran of Korean tech. The launch came stacked with deployments: NAVER is rolling Claude Code across its full engineering org, Samsung SDS is putting Claude Cowork and Claude Code into Samsung Electronics, LG CNS is deploying Claude across the LG Group, and Nexon, Hanwha Solutions, and Channel Corp round out the enterprise list.
It’s not just commercial. Anthropic signed an MOU with Korea’s Ministry of Science and ICT on AI safety, model testing, and cyber threats, and will give Claude access to up to 60 researchers across the NAIRL consortium — KAIST, Korea University, Yonsei, and POSTECH. South Korea, the company says, has become one of its most important international markets.
Fable 5 Ban: A Privacy-Policy ID Check May Reopen Access for US Citizens
A path around the export ban is taking shape — inside Anthropic’s legal fine print. Its updated privacy policy, effective July 8, adds the option to ask customers for a government ID. If the directive holds, Anthropic could reopen Fable 5 and Mythos 5 to users who submit proof of US citizenship, such as a passport or enhanced driver’s license, since real-time citizenship segmentation across a user base in the hundreds of millions was impossible on same-day notice.
Meanwhile the diplomacy continues. Anthropic staff are still meeting with White House and Commerce officials to unwind the control, and the company maintains that recalling a shipped model over a narrow, unproven jailbreak would “essentially halt all new model deployments” industry-wide. Both models remain dark for now.
Claude Design Overhaul: Design-System Imports, Code Round-Trips, and a Token Fix
Anthropic gave Claude Design its biggest update yet. Teams can now import one or more design systems — from a GitHub repo, design files, or raw uploads — and Claude builds with those components, checks its output against them, and auto-corrects before you ever see it. A new admin role can lock a single approved system so every asset stays on brand, and a refined image editor adds fine-grained control over positioning and alignment.
The headline for builders: round-trips with Claude Code. When a design is ready, it hands off to Claude Code — no screenshot, no rebuild — and a /design command lets devs create, edit, and sync projects from the terminal. Crucially, Claude Design now shares usage limits with Code and chat, a direct answer to reviewers who torched most of a weekly Pro allowance generating a few page variants.
The June 15 Agent SDK Billing Split Is Paused — Nothing Changes for Now
Good news if you build agents on a subscription: the change that would have pulled Claude Agent SDK and claude -p headless usage out of Pro, Max, Team, and Enterprise pools — replacing the flat-rate subsidy with a separate, no-rollover dollar credit — is not taking effect. Anthropic confirmed in its Help Center and a note to subscribers that the June 15 move is off.
Those surfaces keep drawing from your existing subscription limits exactly as before. Anthropic says it’s reworking the plan “to better support how users build with Claude subscriptions” and will give advance notice before any future change lands. Translation: don’t re-architect your billing around the old memo — but keep an eye out for the next one.
Claude Drops Into Apple’s Foundation Models Framework
Anthropic shipped ClaudeForFoundationModels, a Swift package that makes Claude a drop-in provider inside Apple’s Foundation Models framework on iOS 27, iPadOS 27, macOS 27, visionOS 27, and watchOS 27. Prototype on Apple’s on-device model, then route complex queries to Claude by swapping a Swift Package Manager dependency — no changes to your session logic.
The package handles streaming, tool calls, and structured responses back into SwiftUI. Requests go straight from the app to the Claude API — Apple never sees the prompts or responses — and usage bills to your Anthropic account at standard API rates. It’s a clean on-ramp for the millions of developers already living in Xcode.
Claude Corps: A $150M Fellowship Puts AI Talent Inside Nonprofits
Anthropic is budgeting $150 million to train young technologists and nonprofit staff to deploy Claude in the field. Claude Corps starts with 100 fellows and aims to scale to 1,000 across as many as 400 nonprofits. Fellows get an $85,000 salary plus benefits, a CodePath mentor, Anthropic office hours, and a generous token budget; host orgs get a $10,000 implementation grant.
The terms are wide open: anyone over 18 with under two years of full-time experience can apply, regardless of degree, as long as they’re US work-authorized and willing to relocate. Applications close July 17 with an October 19 start. Partners include CodePath and Social Finance — a notably hands-on bet on getting Claude into mission-driven workflows.
Anthropic Goes Global Exactly as Washington Pulls It Back
The Seoul launch and the Fable 5 ban are the same story told from two ends. One week the US government freezes Anthropic’s frontier tier over a contested jailbreak; the next, Anthropic plants a flag in Seoul with NAVER, Samsung, and a government safety MOU. The international push isn’t a coincidence — it’s a hedge. When your home regulator can dark a shipped model on same-day notice, geographic and customer diversification stops being growth strategy and starts being risk management.
There’s an IPO subtext too. A company reportedly eyeing public markets wants a revenue base that doesn’t wobble every time an export directive lands. Korean enterprise logos and a research consortium are exactly the kind of durable, in-region demand that reads well to investors — and that no single White House memo can switch off.
Shared Usage Limits Are the Real Headline in Claude Design
The design-system imports are slick, but the quiet line — Claude Design now shares usage limits with Code and chat — is what actually unblocks adoption. A tool that ate most of a weekly allowance in 25 minutes was never going to become a daily habit, no matter how good the canvas. Pricing friction kills products that demos can’t.
The strategic move underneath is the round-trip with Claude Code. Anthropic is stitching design and code into one continuous loop inside its own surfaces, so the handoff that usually leaks time (and tools) never leaves the building. That’s less “AI design app” and more “design-to-ship pipeline” — and it’s a harder thing for point tools to compete with.