Anthropic this week launched Claude for Financial Services, a packaged offering aimed at banks, insurers and asset managers. The release combines a fine-tuned Claude variant with reference architectures for research, KYC review and regulatory drafting, plus deployment options on AWS Bedrock and Google Cloud Vertex AI.

What sets the announcement apart is the level of disclosure. Anthropic published a 40-page model card covering training-data provenance categories, refusal behaviour on market-moving questions, and benchmark results against FINRA and FCA-aligned evaluation suites built with three external auditors. The company also committed to a 90-day notice period before any material behaviour change in the financial variant.

Anthropic says early customers include several tier-one investment banks running internal pilots for equity research and credit memo drafting. The financial variant is priced at a premium over the standard Claude API and is available only under an enterprise agreement with indemnification.