The Legal AI Assessment Framework
A stage-gated framework to evaluate and select GenAI legal assistants. Built by practitioners, open-source, and free for every legal team.
Legal teams have well-established playbooks for hiring humans, but none for hiring AI agents. We're building that playbook together with the legal industry.
Waitlist
Be the first to know when the guide is released.
The v1 Assessment Guide launches in early March 2026. Join the waitlist for early access and contributor updates.
About the Project
Why we're building this
Most legal teams are reinventing the wheel. Custom vendor questionnaires, ad-hoc pilots, word-of-mouth recommendations. There is no shared standard, so every team duplicates effort, procurement quality varies wildly, and vendors and buyers can't even agree on what “good” looks like.
Legal Benchmarks is a fully open-source initiative that helps legal professionals make better-informed decisions when evaluating AI tools. Everything we build is publicly available and shaped by the people who use it. The idea is straightforward: no legal team should have to figure this out alone.
The framework offers a structured, practical approach to evaluating third-party AI vendors, from early shortlisting through pilot assessments to high-risk or cross-border deployments that require deeper due diligence. It is tailored to GenAI legal assistants that access internal company data, are used for core legal work, and operate across the full lifecycle: retrieve, generate, store, act.
Independent
No vendor sponsorship or pay-for-placement.
Practitioner-led
Shaped by in-house lawyers and legal practitioners.
Open-source
Publicly available and community-driven.
The Team
Core Working Group
Founder, Legal Benchmarks
Over the past 13 months, Anna has benchmarked AI performance in real legal workflows alongside 500+ legal professionals. That work has shaped hundreds of procurement decisions around AI tooling.
LinkedIn
Technologist & Former Lawyer
Led legal AI procurement and evaluated 40+ tools last year. Brings an end-to-end view of AI assessment, from screening and testing to piloting and making the final tool decision.
LinkedIn
General Counsel
Focused on data security and operational risk in AI systems. Pushes the team to ask the questions most organisations miss around security, safety, and deployment risk.
LinkedInAdvisory Board
Advisory Board
A small group of senior legal leaders (GCs, CLOs, and decision-makers from in-house and private practice) who guide strategic decisions, share their perspective on new directions, and review drafts before they go public.
Advisory board members will be announced as they are confirmed.
Steering Committee
Steering Committee
A broader group of practitioners who validate and shape the assessment criteria, review and give feedback on the framework via survey or direct comments, help recruit others in their network, and publicly signal support. All members are acknowledged in the final published v1 guideline and toolkit.
Steering committee members will be listed as they are onboarded. Apply to join.
Roadmap
Timeline
Framework Development
Synthesised legal AI evaluation frameworks used in real vendor selections. First draft assembled November to December 2025.
Community Feedback
Draft opened for community input. 50+ legal teams have contributed so far. The steering committee and advisory board are reviewing the updated draft via survey and Google Doc comments.
v1 Assessment Guide Finalisation
Late Feb 2026Incorporating community input into the v1 Assessment Guide and Toolkit, including the pre-demo checklist, demo scorecard, pilot assessment scorecard, and vendor comparison template.
Publication and Launch
Early March 2026Public release of the Legal AI Assessment Framework v1, the Assessment Kit, and contributor credits. Launch announcement via newsletter, LinkedIn, and partner channels.
FAQ
Frequently Asked Questions
Get Involved
Join the steering committee
We are looking for in-house counsel, legal-ops professionals, law firm innovation leads, academics, and legal technologists. The commitment is light: review the draft framework and complete a short survey (about 15 minutes). All contributors are acknowledged in the published v1 guideline and toolkit.