Benchmarking Generative AI for CSA Clause Extraction and CDM Representation

This whitepaper explores recent advances in generative artificial intelligence (AI) and its potential to automate the extraction, interpretation and digitization of legal clauses from ISDA’s credit support annexes (CSAs) into standardized, machine-readable code using the Common Domain Model (CDM). Specifically, the paper benchmarks and reviews the ability of generative AI to accurately extract five CSA clauses and digitize them into CDM format. The findings suggest that integrating industry-specific data significantly boosts generative AI accuracy, larger generative AI models typically handle nuanced legal language better, and generative AI can be utilized within a modular framework (eg,agentic AI) to extract legal clauses more accurately and efficiently.

Click on the attached PDF to read the full report.

Documents (1) for Benchmarking Generative AI for CSA Clause Extraction and CDM Representation

Pursuing Efficiencies of Tokenization

As we mark the 10-year anniversary of the ISDA Standard Initial Margin Model (ISDA SIMM), which was launched in September 2016 to coincide with the first phase of initial margin (IM) requirements for non-cleared derivatives, we’ve been reflecting on its...

Transition to Mandatory Central Clearing

US Treasury securities sit at the heart of global financial markets and serve as one of the primary forms of high-quality collateral across derivatives and securities financing markets. The transition to mandatory central clearing of US Treasuries therefore has implications...