SAP and Anthropic Expand: Claude Becomes the Reasoning Layer for SAP’s Business AI Platform
At SAP Sapphire, SAP and Anthropic announced a major expansion of their partnership: Claude will be embedded as the primary reasoning and agentic capability across SAP’s AI-enabled solution portfolio and will power the newly announced SAP Business AI Platform. The platform is SAP’s bet for the agentic-enterprise era — a unified place where customers build, run, and govern AI agents on top of SAP data and processes.
For Anthropic, it’s a foothold inside the systems that run the world’s largest companies. SAP’s installed base spans finance, supply chain, HR, and procurement — the workflows enterprise AI has been trying to crack for two years. Pairing Claude’s reasoning with SAP’s data plane gives the partnership a credible answer to Microsoft Copilot in the enterprise stack, and it lands on the same day the KPMG and Bristol Myers Squibb deals are still echoing through the trade press.
Bristol Myers Squibb Adopts Claude Enterprise as Shared Intelligence Platform
Bristol Myers Squibb announced a strategic agreement with Anthropic to deploy Claude Enterprise across research, clinical development, manufacturing, commercial, and corporate functions — effectively company-wide. BMS framed Claude as a “shared intelligence platform” that connects scientists, clinical teams, and operators on a common reasoning layer, with rollouts already underway.
It’s a marquee pharma win in a vertical that’s been cautious about generative AI. Drug-discovery teams have specific concerns — IP isolation, auditability, regulated workflows — that match cleanly with Anthropic’s recent enterprise push (managed agents, self-hosted sandboxes, MCP tunnels). Expect BMS to become the case study Anthropic points at when selling into the rest of pharma.
Anthropic Opens Seoul Office; Names KiYoung Choi Korea Representative Director
Anthropic appointed KiYoung Choi as Representative Director of Korea ahead of a Seoul office opening, the company’s latest international expansion after Tokyo and Dublin. Korea is a strategic AI market — home to Samsung, SK Hynix, and a deep base of enterprise software buyers — and one where local presence routinely matters for enterprise procurement.
The hire signals serious commercial intent rather than a symbolic flag-plant. Watch for Korean-language Claude support and partnerships with Korean cloud providers and chaebols in the coming quarters; the playbook from Tokyo and London suggests a wave of vertical deals tends to follow within six months of a regional GM landing.
28 New Security and Compliance Integrations for Enterprise Claude Governance
Anthropic shipped 28 new security and compliance tool integrations to help IT and security teams govern Claude inside their existing stacks. The integrations span DLP, SASE, data security, SIEM, security operations, identity management, eDiscovery, AI security posture management, and observability — the categories every enterprise security architect has on their diagram.
The plain-English read: CISOs no longer have to choose between blocking Claude and rolling it out blind. They can route Claude traffic through their existing DLP, see Claude prompts and responses in their SIEM, manage Claude access through their identity provider, and pull eDiscovery exports when legal asks. It’s the kind of unsexy plumbing that decides whether a year-long pilot turns into a company-wide rollout, and it’s what the Big Four and pharma deals announced this week were quietly waiting for.
Claude Platform on AWS Goes Generally Available
AWS made the Claude Platform generally available, the first cloud provider to offer Anthropic’s native developer surface from inside an existing AWS account. Customers can now use the Claude APIs, the Claude Console, and early-access beta features without setting up a separate Anthropic account — billing, identity, and procurement stay inside AWS.
For AWS-standardized shops, that’s a real friction killer: no new vendor onboarding, no parallel SSO, no separate invoice. It also reinforces the AWS–Anthropic alignment at a time when Microsoft is leaning into its OpenAI integration. With Claude Opus 4.7 on Bedrock and the Claude Platform now first-class on AWS, builders in the AWS ecosystem have one of the smoothest paths to production-grade Claude available anywhere.
Claude Code Picks Up Security Plugin and Faster Performance
The Claude Code team shipped a fresh round of updates. A new security-guidance plugin, now available to every Claude Code user, monitors code edits, diffs, and commits in real time and uses regex-based pattern matching to flag roughly 25 high-risk vulnerability classes — SQL injection, command injection, XSS, hardcoded API keys, improper input validation. It runs as a default-on safety net rather than a separate audit tool.
Alongside the plugin, the team posted broader performance and reliability gains: smarter background and pinned sessions, a renamed /code-review flow, fixes across terminal, Windows, PowerShell, plugin, and slash-command issues, and built-in type-to-filter in the new /skills and /plugin prompts. Less glamorous than a model release, but the kind of polish that keeps daily users on Claude Code.
Base Launches MCP Gateway: Claude Gets Onchain DeFi Actions
Base, Coinbase’s Ethereum Layer-2, launched Base MCP this week — an MCP gateway that lets AI agents like Claude execute onchain DeFi actions through natural language. The gateway wires Claude into six protocols at launch, including Uniswap, Morpho, and Avantis, and routes transactions without ever exposing user private keys to the model.
It’s an early but concrete look at what agentic crypto could feel like: ask Claude to rebalance a portfolio or open a position and the gateway handles the signing. Plenty of caveats — this is DeFi, the safety surface is wider than typical SaaS — but it’s another sign that MCP is colonizing the integration layer well beyond enterprise SaaS.
MCP Hits 2,300+ Public Servers; Ecosystem Crosses 200 Tools
The Model Context Protocol has crossed 2,300 public servers, with adoption now spanning Claude, Cursor, Windsurf, VS Code, and more than 200 other tools. Anthropic donated MCP to the Linux Foundation in December 2025 under the new Agentic AI Foundation co-founded with Block and OpenAI — an unusually neutral home for a protocol whose original sponsor is suddenly its biggest commercial beneficiary.
For developers, the practical signal is that MCP is no longer a Claude-specific contract. Build a server once and it works across most frontier-model clients, which is exactly the kind of network effect that turns a protocol into infrastructure.
The Enterprise Land Grab Is the Real Story of Q2
Step back from the daily noise and look at May’s deal flow as a single document. KPMG (276,000+ seats). PwC (a 30,000-person training program). SAP (embedded across the Business AI Platform). Bristol Myers Squibb (company-wide). Salesforce (regulated industries). Thomson Reuters (CoCounsel). AWS (native Claude Platform). Add 28 security integrations, MCP tunnels, self-hosted sandboxes, and a Korea office opening, and the picture is unmistakable: Anthropic isn’t trying to win consumer; it’s racing to become the default reasoning layer in regulated, large-enterprise software.
The strategic logic is tidy. Enterprise revenue is sticky, multi-year, and high-margin; it doesn’t depend on ads or viral consumer wins; and once Claude is wired into SAP and into a Big Four delivery practice, switching costs balloon. The risk is the same one OpenAI has been managing: when your biggest customers are also strategic partners (and sometimes competitors), product roadmap conversations get complicated fast. For now, the math is on Anthropic’s side — ~$30B ARR, 80× growth in programmatic usage, and a list of marquee logos most rivals would trade quarters for. Q2 is shaping up as the enterprise land-grab quarter, and Anthropic is moving fastest.