Benchmarks from NextLadder Ventures and Vals AI, built with legal aid experts, tested 23 AI models on 252 legal scenarios often faced by low-income Americans

Analysis found general-purpose AI models often provide generic legal information without asking key questions or identifying available local help – critical metrics for the next generation of tools to improve upon.

SAN FRANCISCO, Nov. 10, 2026 /PRNewswire/ — Today, Vals AI and NextLadder Ventures released the Access to Justice Benchmark, an evaluation of 23 AI models across 252 legal scenarios written by legal aid experts in six states. Models met 78–87% of criteria for general explanations of the law, but only 30–50% for identifying specific resources, such as local legal aid programs or court self-help centers.

The benchmark is part of a broader effort spearheaded by NextLadder Ventures to accelerate the development of AI tools tailored to the unique needs and challenges facing low-income Americans navigating fork-in-the-road moments with disproportionate impact on their economic mobility.

“AI could help millions of people facing a legal problem from housing to family law understand their options and get connected to the right help,” said Ryan Rippel, CEO of NextLadder Ventures. “Getting there means knowing where today’s tools can be trusted and where they fall short. The Access to Justice benchmark gives us that map, showing exactly where AI models need to get better and where specialized tools built with legal aid experts can make the biggest difference in the future. We hope that independent builders and the frontier labs will both take up this challenge to build tools that can help Americans who too often cannot afford the civil legal help they and their families need.”

Most low-income Americans face their civil legal problems without a lawyer. According to the Legal Services Corporation’s 2022 Justice Gap study, low-income Americans did not get any or enough legal help for 92% of their civil legal problems, and nearly half (46%) of those who didn’t seek help cited cost as a reason. Trusted, purpose-built AI-powered tools could help close this gap and increase the bandwidth of human legal aid experts and attorneys to help families in need at key moments like an eviction notice, tenant’s rights issue, child support fight, or domestic violence protective filing.

“Most legal AI benchmarks measure work done for law firms and corporate clients,” said Meghana Kotcherlakota, Research Scientist at Vals AI. “We built the Access to Justice Benchmark to show how well these models serve the people who need help most, problem by problem, and to track whether they get better.”

The results of the Access to Justice Benchmark, which tested 23 leading AI models on 252 real-world scenarios written by legal aid experts in six states, are posted HERE. Key takeaways include:

• Models miss key steps in helping people navigate legal problems. Even the top model, Qwen 3.8 Max (76.4%), misses about a quarter of what experts expect, like asking the right questions to understand someone’s context before giving guidance. The top three are within the margin of error of each other, and the top two OpenAI models trail at about 66%.
• Models show wide variance in effectiveness across topics. Models do comparatively better on housing and landlord-tenant questions (75.4% across all models). Public benefits questions rank worst (57.3% across all models). Family law scores varied widely by subtopic: visitation (61.3%) and kinship caregiving with child protective services (63.0%) score lowest, and guardianship, divorce finances, family-member adoptions, and domestic violence fall between 66% and 68%.
• Models perform better on legal explanations than on referrals to local help. Every model meets 78–87% of criteria for general explanations of the law, but only 30–50% for specific resources such as a local legal aid program or court self-help center. For someone navigating a legal problem without a lawyer, practical next steps – including relevant legal aid services and court resources – are an important part of a useful answer.
• Premium models show little advantage over free-tier models in this test. Free-tier chatbot models average 69.7% and premium ones 70.4%. Anthropic’s free-tier Claude Sonnet 5.5 scores 1.2 percentage points higher than Claude Opus 5.5. Only the xAI gap, 2.3 percentage points in favor of premium Grok 4.7, is larger than the margin of error.

Vals AI and NextLadder Ventures are collaborating on a full report and an additional brief to help frontline practitioners understand the gaps that remain to be closed with these tools. 

Examples from the Benchmark Tests

Louisiana small claims
In one test, a parent in New Orleans asked whether they could take their child’s other parent to small claims court over money promised verbally and never repaid. Most models got the basics right: they pointed to First City Court (or Second City Court for Algiers), cited Louisiana’s 10-year prescriptive period and said a witness would be needed without a written contract. But only 8 of 23 asked whether the debt involved real estate, which decides whether a verbal agreement can be enforced at all. Legal aid experts also expected them to warn that the claim is lost once the 10 years run out (4 of 23 did) and to send the client to the court clerk’s website for the Statement of Claim and fee-waiver forms (3 of 23 did).

NYC constructive eviction and deposit
In another, a New York City tenant who had moved out after a constructive eviction asked how to recover a deposit the landlord never returned. Nearly every model cited the landlord’s 14-day deadline to return or itemize the deposit and laid out a demand letter followed by a small claims filing. 14 of 23 were marked down for pushing that out-of-court step, which the expert’s rubric warns against. Experts expected the model to refer the client to NYC’s Right to Counsel program, and only 4 of 23 did. Most gave general hotlines like 311 or Legal Services NYC instead.

About NextLadder Ventures: NextLadder Ventures is a time-bound venture with one goal: empower millions of Americans to reach their potential by 2040. Backed by over $1 billion in capital, the organization invests in breakthrough technologies that remove barriers to economic success and put people in control of their futures. NextLadder Ventures is trailblazing a new market for tech-enabled Navigation Technology tools that help people access the resources they need to navigate pivotal moments — offering flexible, risk-tolerant capital to entrepreneurs building these transformative tools today, while creating a pipeline of tech, talent, and capital for the long run. To learn more, visit nextladder.com.

About Vals AI: Vals AI is an independent, third-party AI benchmarking and evaluation company, and the standard used by every frontier AI lab and cited in their model releases. Vals designs and operates rigorous benchmarks that measure how AI models perform on real-world professional tasks across domains including cyber, military, legal, finance, healthcare, coding, public benefits, and youth safety. Its work has been covered by outlets including the New York Times and Washington Post. Vals is a Series A company backed by a16z and headquartered in San Francisco.

View original content:https://www.prnewswire.com/news-releases/new-test-of-leading-ai-models-legal-aid-advice-finds-key-gaps-still-to-be-closed-across-issues-commonly-faced-by-low-income-americans-302935069.html

SOURCE NextLadder Ventures