DEV Community

Cover image for OpenAI and Sullivan & Cromwell Build AI Agreement Analyzer
AI Feed
AI Feed

Posted on Originally published at aifeed.space

OpenAI and Sullivan & Cromwell Build AI Agreement Analyzer

Agreement Analyzer reviews draft contracts against selected precedents and negotiating positions while leaving every legal decision to lawyers.

OpenAI said on October 15 that Sullivan & Cromwell, working with the AI company, developed Agreement Analyzer to help the law firm’s lawyers identify potential issues in draft agreements. The application combines the firm’s legal materials and established workflows with ChatGPT Enterprise, but does not make final legal decisions.

The project began as a single-user proof of concept built with Codex by Sullivan & Cromwell partner John Savva. OpenAI subsequently worked with the firm to improve output quality, introduce parallel processing and enable several lawyers to use the application simultaneously.

For a review, a lawyer can provide a draft agreement alongside selected precedents, firm and client negotiating positions, and information about market practice. According to OpenAI, Agreement Analyzer examines the material through several review cycles from different perspectives, consolidates the results and flags potential problems. It then produces a memorandum and proposed edits presented as tracked changes.

OpenAI says specialized agents examine deviations from market practice, the allocation of risk, relationships among clauses and ambiguities that might affect litigation. The agents can challenge one another’s conclusions. The resulting findings are separated into those that survive the challenge, those that are dismissed and unresolved matters retained with qualifications.

The company explicitly describes Agreement Analyzer as an aid rather than a substitute for legal expertise. Lawyers can inspect its explanations and supporting precedents, then adapt, reject or accept proposed changes. Its memorandum, findings and tracked edits all remain subject to professional review and judgment.

The approach matters because it aims to expose both the evidence behind a recommendation and the uncertainty surrounding it, rather than presenting every machine-generated finding as equivalent. Categorizing conclusions as sustained, dismissed or unresolved could help lawyers focus their attention on provisions where transaction context and professional judgment are especially important.

The announcement does not provide pricing, a timetable for broader deployment or availability outside Sullivan & Cromwell. OpenAI characterizes wider research and writing assistance as a future opportunity, not a completed product capability. The firm is continuing to explore other tools tailored to specific legal tasks.

Sources

Originally published by AI Feed: OpenAI and Sullivan & Cromwell Build AI Agreement Analyzer.

Top comments (0)