Broadridge Joins Anthropic’s Project Glasswing

Broadridge has joined Anthropic’s Project Glasswing to evaluate the startup’s large language models for use in financial services and back-office operations. The agreement gives Broadridge access to Anthropic’s enterprise AI offering to test model capabilities and governance tools for its client services.

The companies signed onto the program recently. Pilots will run in controlled environments where Broadridge will assess model performance, privacy protections and compliance controls before any wider deployment.

Project Glasswing provides customers with access to Anthropic’s language models along with features intended to support safety and operational controls. Participating firms can run pilots, tune models for domain-specific tasks and apply guardrails designed to reduce risky outputs.

Broadridge plans to evaluate the technology for automating routine client communications, improving document processing and supporting internal workflows. Testing will concentrate on data handling, regulatory requirements and integration with existing systems.

The firms did not disclose financial terms or a public timeline for pilots. Broadridge and Anthropic will define success criteria, accuracy metrics and measures for protecting sensitive customer data as part of the tests.

Broadridge has invested in automation and cloud services and has previously partnered with third parties to speed development. Anthropic, an AI startup focused on safety and alignment, has been offering enterprise access to its models with safety features and monitoring.

Industry observers expect financial services firms to emphasize oversight and explainability as they adopt AI tools, citing strict privacy and regulatory standards. The partnership adds to collaborations between fintech providers and AI developers working on generative AI use in regulated industries under controlled conditions.

No direct quotations were provided by either company at the time of the announcement.