Consensus to Courtroom: Sedona WG13's Roadmap for AI Legal Reform and Practical Guidance
TSC WG13 convened judges, scholars, and technologists to assess AI's benefits, risks, and gaps, aiming for practical tools. Deliverables: definitions, governance, court guidance.
The Sedona Conference Working Group 13: Legal Reform and Practical Guidance for AI's Impact on the Law
On September 11-12, 2025, The Sedona Conference (TSC) Working Group 13 met at the Hyatt Regency in Reston, Virginia for a sold-out Midyear Meeting with more than 135 participants. Judges, legal scholars, practitioners, technologists, ethicists, and policymakers examined how AI is changing legal systems, regulatory structures, and society.
The program mixed broad panels with focused work sessions. The goal: surface risks, identify gaps, and produce practical guidance that courts, practitioners, and legislators can use. The takeaway was clear-AI is here, it's advancing, and the profession needs consensus tools to address it.
Key Benefits Highlighted
- Expanded access to justice: GenAI is helping non-legal professionals prepare higher-quality filings. Patent applications at the USPTO are improving when supported by AI tooling.
- Stronger research and writing: Case law-trained models are moving beyond keyword search to more analytical outputs.
- Fewer hallucinations: Newer models show better reliability, with measurable drops in fabricated citations.
- Firm integration: Structured deployments with guardrails are improving client service without derailing associate training.
Where Caution Is Warranted
Participants warned that overuse can dull core legal reasoning. GenAI predicts; it does not think. The group discussed developing guidance for law schools and practitioners on balancing efficiency with skill development.
Speed and opacity were recurring issues. Products ship before thorough testing, bias persists, and black-box outputs complicate client counseling and admissibility. The room called for transparent evaluation methods before wide release, and clearer explanations of how systems produce results.
Do We Need AI-Specific Laws?
- Use existing law: Some argued that current statutes (e.g., Title VII, ADEA, Equal Pay Act) address many risks and that new laws could create confusion.
- Don't smother innovation: Premature regulation may add compliance cost without improving outcomes.
- Smart harmonization: A jurist urged agencies to use existing authority to issue AI-related guidance (e.g., in IP) instead of leaving coherence to fragmented case law.
- Standards with care: Industry standards can help, but participants cautioned against early certification schemes that inflate cost or create superficial "safety" labels.
Gaps That Need Attention
- Intellectual property: Authorship, ownership, and infringement questions remain unsettled when AI contributes to content creation.
- Tort liability: Accountability for AI-caused harm is unclear; targeted statutes, regulations, or standards may be necessary.
Cross-Cutting Concerns
- Privacy and security: Compliance remains complex across state, federal, and international regimes.
- Court operations: From AI-generated briefs citing fake cases to authenticating AI-produced evidence, lower courts are inconsistent.
- Environmental impact: Data center demand could strain resources without oversight.
Work in Progress: Concrete Outputs from WG13
The meeting focused on four workstreams aimed at delivering practical tools for the legal community.
Practice Guides
- Consensus Definitions: With no common definition of "AI," the group is developing a usable definition for courts, practitioners, and legislators.
- Governance Framework: A practical guide for deployers to integrate AI into organizations with accountability, evaluation, and controls.
- Regulatory Crosswalk: A resource for courts, legislators, and practitioners-either as a unified tool or audience-specific versions.
- Additional resources: Checklists and best-practice manuals, including prompt development guidance for lawyers and clients.
Judicial Guidance
Judges described two evidence categories: Acknowledged AI (disclosed in filings or proposed evidence) and Unacknowledged AI (forged evidence and deepfakes). Standardized approaches to authentication and disclosure would help, but rigid rules could age poorly.
TSC will consider a subgroup focused on court guidance for deepfakes and related issues, including authentication practices and standing orders. For context on authentication principles, see Federal Rule of Evidence 901.
Cross-Disciplinary Collaboration
Participants agreed that sustained collaboration across legal, technical, policy, and ethics communities is essential. The meeting validated ongoing efforts and surfaced the need for additional drafting committees as new challenges appear.
What Legal Teams Can Do Now
- Set policy and training: Establish AI use policies covering disclosure, confidentiality, and human review. Train teams on model limitations and verification.
- Adopt evaluation protocols: Require testing for bias, reliability, and data provenance before deployment. Document results and mitigation steps. Consider frameworks like the NIST AI Risk Management Framework.
- Strengthen authentication: Update litigation hold notices, ESI protocols, and evidence checklists to account for AI-generated content and deepfakes.
- Protect core skills: Pair AI tools with deliberate practice: issue spotting, statutory interpretation, and fact development.
- Monitor regulation: Track agency guidance and judicial trends; maintain a living crosswalk of obligations across jurisdictions.
- Engage clients: For clients deploying AI, implement governance: inventories, risk ratings, role-based access, and escalation paths.
Conclusion
AI offers clear benefits for legal practice, but it also introduces risk, uncertainty, and uneven adoption across courts and agencies. TSC's WG13 is moving from issue-spotting to practical outputs-definitions, governance guidance, crosswalks, and judicial resources-that legal professionals can apply now.
If your team is building prompt discipline and needs structured practice, see this resource: Prompt Courses at Complete AI Training.
