Dun & Bradstreet Integrates Risk Data into Anthropic Claude

Dun & Bradstreet announced a partnership with Anthropic to integrate its commercial risk dataset, the Commercial Graph, into Anthropic's Claude, per a Dun & Bradstreet press release and coverage by PYMNTS and Dealroom. The companies said the connection is implemented through a Model Context Protocol (MCP) server, which supplies Claude with verified commercial context anchored by the D-U-N-S Number, the global business identifier cited in the release. The integration is framed as enabling automated know-your-customer (KYC) and know-your-business (KYB) onboarding, with specific capabilities named in the announcement including entity identity verification, ownership-structure assessment, third-party network analysis and generation of audit-ready documentation, according to Dealroom and the press release. PYMNTS reproduced a direct quote from Dun & Bradstreet's announcement where the company's general manager of risk, Zuck, described the system as providing "verified context and decision logic."

Per the companies' announcement and coverage, the technical architecture centers on exposing D&B's Commercial Graph to Claude through an MCP server that delivers structured, authoritative context at inference time. The sources describe the integration as allowing prompt-driven agents inside Claude to combine natural-language instructions with real-time, verified business data. Reported capabilities include:

• •verifying business identities using the D-U-N-S Number
• •resolving ownership and control structures
• •evaluating third-party risk networks
• •producing audit-ready documentation for compliance records

Editorial analysis: Companies integrating verified enterprise data with large language models are addressing two recurring problems in regulated automation: the need for authoritative ground truth and the need for auditable reasoning trails. Industry reporting places this partnership alongside other Anthropic commercial deals that embed structured data into model context to support high-assurance workflows in banking and FinTech, per Jacksonville Business Journal and PYMNTS. Observed patterns in similar integrations show organizations trade off some model flexibility for stronger governance, typically by routing high-stakes lookups to trusted data sources and attaching provenance metadata to outputs.

Editorial analysis: For practitioners building compliance and onboarding systems, the deal illustrates a maturing integration pattern: combine an LLM agent interface with verified domain data served at inference time. This reduces reliance on brittle prompt engineering and post-hoc reconciliation while making it easier to construct deterministic checks around model outputs. The use of persistent identifiers such as the D-U-N-S Number is particularly valuable in enterprise settings because it enables consistent joins and traceability across datasets.

observers should track:

• •whether the integration exposes provenance metadata and confidence scores in production outputs
• •how access controls and consent are implemented between the MCP layer and enterprise clients
• •early adopter case studies reporting cycle time reduction or audit findings, which sources say are target benefits. Also watch regulatory scrutiny or industry standards updates that address machine-assisted decisioning and recordkeeping, since those will affect deployment scope in banking and other regulated sectors