🧭 Anthropic Opens Seoul Office: MOU with Korea's Ministry of Science, NAVER, Samsung SDS and More

Anthropic formally opened its Seoul office — its third Asia-Pacific location after Tokyo and Bengaluru — and announced a dense web of partnerships spanning government, enterprise, academia, and startups. KiYoung Choi, a 30-year technology leader and former Korea GM at Snowflake, joins as Representative Director. The centrepiece is a Memorandum of Understanding with Korea's Ministry of Science and ICT covering AI safety and cybersecurity collaboration, paired with a separate agreement with the Korea AI Safety Institute to develop Korean-language model evaluation frameworks — a direct complement to Anthropic's existing commitments with the UK and US AI Safety Institutes.

Enterprise partnerships

The enterprise partner list is one of the broadest Anthropic has announced at a single regional launch:

• NAVER: Deploying Claude Code across its entire engineering organisation — the first Korean internet company to adopt Claude Code at org-wide scale.
• Samsung SDS: Rolling out both Claude Cowork and Claude Code across Samsung Electronics, giving the conglomerate's developers and knowledge workers access to Claude in their existing workflows.
• LG CNS: Deploying Claude to thousands of employees across LG's IT services arm.
• Nexon: Using Claude Code for live-service game development — the first major Korean games studio to commit publicly to agentic coding at this scale.
• Hanwha Solutions: Accessing Claude via AWS Bedrock with Korean data residency — important for a company in the energy and materials sectors that operates under strict data governance.
• Channel Corp: Powering Channel Talk — a B2B customer communication platform used by over 230,000 businesses — with Claude.

Academic and developer ecosystem

The National AI Research Lab, a consortium spanning KAIST, Korea University, Yonsei, and POSTECH, will give approximately 60 researchers access to Claude for frontier AI research. On the developer side, Claude for Startups is now available in Korea; Anthropic also ran a Claude Build Day with BASS Ventures and a Push to Prod hackathon co-hosted with Replit and Korea Investment Partners.

🧭 Workload Identity Federation Is Generally Available: Replace Static API Keys with Keyless Auth

Anthropic's Workload Identity Federation (WIF) has reached general availability across all Claude API endpoints, allowing any workload to authenticate to the Claude API using its existing cloud identity — no static API keys required. WIF works by having the Claude API trust a set of configured identity providers: your workload presents its existing short-lived credential (an IAM role token, a Kubernetes service account JWT, a GitHub Actions OIDC token, etc.), and the API exchanges it for a session-scoped Claude API credential. The static API key never exists and therefore cannot leak.

Supported identity providers at GA

• AWS IAM roles — EKS pods, Lambda functions, EC2 instance profiles
• Google Cloud — GKE Workload Identity, Cloud Run service accounts, Vertex AI pipelines
• Azure Managed Identities — both system-assigned and user-assigned
• Kubernetes — service account token projection (any OIDC-compliant cluster)
• GitHub Actions — per-job OIDC tokens (distinct identity per workflow run)
• Okta and other OIDC-compliant enterprise IdPs

Service accounts

WIF launches alongside service accounts in Claude Console: create a named service account per workload (e.g. prod-data-pipeline, staging-code-review), assign it a role and rate-limit profile, and bind it to one or more trusted identity provider subjects. Each workload then gets its own distinct identity in the API audit log — making it easy to attribute costs, enforce per-workload spend limits, and revoke access for a single pipeline without rotating a shared key that other systems depend on.

# Example: bind a GitHub Actions workflow to a Claude service account
# (in Claude Console → Service Accounts → OIDC Configuration)
{
  "provider": "https://token.actions.githubusercontent.com",
  "subject": "repo:my-org/my-repo:ref:refs/heads/main",
  "audience": "https://api.anthropic.com"
}

🧭 Claude Code Artifacts: Turn Sessions into Shareable Live Web Pages

Claude Code is shipping Artifacts in beta for Team and Enterprise subscribers — a feature that turns the output of any Claude Code session into a persistent, auto-updating web page at a stable URL. Where previously a session's results existed only inside the terminal or in files on disk, Artifacts packages them into a shareable, version-tracked, web-accessible document that updates automatically each time the underlying session produces new content.

What Artifacts are designed for

The feature targets the communication gap between the developer running a session and everyone else who needs to understand what it produced:

• PR walkthroughs: A living document that narrates the changes in a pull request — automatically updated as Claude makes more changes, so reviewers always see the current state without re-reading a thread of comments.
• Incident pages: A shared status page generated from an active debugging session. As Claude Code investigates, the Artifact updates with findings, hypotheses ruled out, and the current best theory — giving on-call engineers and stakeholders a single place to watch progress.
• Dashboards and checklists: Structured summaries of long-running agent sessions — particularly useful when nested sub-agents are running (a feature that shipped June 18) and you want a readable roll-up of what each branch of the agent tree is doing.

Key details

• Version history: Every update to an Artifact is snapshotted, so you can link to any prior state or diff between snapshots — useful for auditing what Claude reported at a specific point in time.
• Sharing controls: Default is org-only (anyone in your Claude Team/Enterprise workspace can view); public sharing requires an explicit toggle. Enterprise orgs can disable public sharing globally via the admin console.
• Compliance API: Enterprises with audit log requirements can pull all Artifact creation, update, and access events via the existing audit log API — the same stream that covers conversations and API calls.