NYU Researchers Develop AI Tool to Identify Risky and Unenforceable Contract Clauses

Computer scientists at New York University have developed an AI tool called ContractNerd designed to identify risky, ambiguous, or unenforceable clauses in contracts, particularly those used by employers and landlords. These contracts often contain terms that place unfair burdens on tenants and employees due to vague language or legally invalid provisions. For example, a lease clause stating "Tenant must provide written notice of intent to vacate at a reasonable time" is problematic because "reasonable" lacks a clear definition. Similarly, broad non-compete clauses like "Employee agrees not to work for any business in the United States for two years following termination" are often unenforceable in many states. ContractNerd uses large language models to analyze contracts and categorize clauses into four types: missing clauses, unenforceable clauses, legally sound clauses, and legally questionable but risky clauses—classified as high, medium, or low risk. The tool focuses on employment and rental agreements in New York City and Chicago, drawing insights from legal databases such as Thomson Reuters Westlaw, Justia, and Agile Legal, while also incorporating state-specific regulations. To assess its performance, researchers compared ContractNerd against existing AI contract analysis tools. In one test, it outperformed others in predicting which clauses would be deemed unenforceable in court. In a second evaluation, a panel of laypeople reviewed outputs from ContractNerd and another leading system, goHeather, without knowing which tool produced which results. ContractNerd received higher ratings overall. In a third assessment, NYU School of Law Professor Clayton Gillette evaluated both tools using real-world clauses, from simple ones like "No pets allowed" to complex ones involving attorney fee obligations. He found ContractNerd more thorough in its analysis, though goHeather’s explanations were easier to understand. The researchers believe ContractNerd can serve as a valuable aid for both contract drafters and signers, helping users identify potential legal risks and promote fairness. "Many of us have to read and decide whether to sign contracts, but few have legal training," said Dennis Shasha, Silver Professor of Computer Science at NYU’s Courant Institute and senior author of the study. "ContractNerd helps identify clauses that are missing, biased, illegal, or ambiguous—and suggests improvements." The team plans to expand the tool’s reach beyond New York and Chicago. The research was published in the journal Electronics, with co-authors Musonda Sinkala and Yuge Duan, NYU graduate students at the time, and Haowen Yuan, an NYU undergraduate.