Policy Paper on the Use of AI Tools by Legal Professionals in Emerging Companies and Venture Capital

Click HERE to access the full report.

SUMMARY

• AI suitability depends on document characteristics, not on the technology. The report identifies five factors — standardization, complexity, tailoring, an organization’s experience with the document type, and the level of negotiation involved — and applies them in an illustrative table. Basic C-corporation formation documents, board minutes, and non-disclosure agreements rate high for AI suitability; shareholder agreements, financing documents, and merger and acquisition agreements rate low, with M&A requiring the highest level of human legal judgment.
• Courts have already sanctioned lawyers for unverified AI output. The report cites Mata v. Avianca, Inc., in which a federal judge sanctioned a lawyer for submitting a brief containing hallucinated ChatGPT citations, and Wadsworth v. Walmart Inc., in which Morgan & Morgan lawyers were sanctioned in February 2025 for a motion containing eight fabricated citations generated by the firm’s in-house AI platform. The report emphasizes the court’s holding that the duty to verify legal authority is personal and cannot be delegated.
• Existing ethical duties already govern AI use. The report maps AI-specific obligations onto MRPC Rules on competence (Rule 1.1 and Comment 8), confidentiality (Rule 1.6), supervision (Rules 5.1 and 5.3), candor to the tribunal (Rule 3.3), client communication (Rule 1.4), advertising (Rule 7.1), fees (Rule 1.5, including a caution against billing time not actually spent), and anti-discrimination (Rule 8.4(g)). It draws throughout on the New York City Bar Association’s Formal Opinion 2024-5, which addresses confidentiality, competence, supervision, candor, and billing in the generative AI context.
• The regulatory landscape is fragmented across jurisdictions. An appendix catalogs official ethics guidance, task force reports, or court guidance from twenty-one jurisdictions, and the report documents federal standing orders — including the first, issued by Judge Brantley Starr of the Northern District of Texas on May 30, 2023 — that variously require certification, mandate disclosure, or prohibit AI-assisted drafting outright.
• The report calls for harmonization and for firm-level governance now. It concludes that the profession would benefit from model guidelines or a national framework built on the MRPC, preserving state interpretive authority while establishing baseline obligations. In the interim, it recommends that practitioners adopt firm-wide protocols and risk assessments, train lawyers and staff, disclose AI use to clients where appropriate, monitor standing orders in relevant jurisdictions, and designate a technology ethics officer — and that law schools embed AI ethics into their curricula.

• Introduction — the current state of AI adoption in legal services and the absence of a uniform regulatory framework.
• How AI tools can be useful to lawyers but could impose risks on non-lawyers — the five-factor suitability framework, the document-by-document suitability table, appropriate use cases (template generation, clause suggestion, contract review and comparison, data extraction, research support, language optimization), and the risks of relying on AI without an attorney.
• Leveraging large language models in venture financing — high-volume standardized agreements, due diligence support, early-stage financing instruments (SAFEs, convertible notes, revenue-based financing), advanced and complex agreements, and current tools in the market.
• Regulation of the use of AI in legal practice — malpractice, data privacy and security, deepfakes and AI-generated misinformation, hallucinations, model bias, and intellectual property; followed by a duty-by-duty analysis under the MRPC.
• Future direction and considerations — the case for harmonization, the role of the ABA and professional associations, anticipated state-level guidance, judicial influence, AI and access to justice, and recommendations for practitioners and institutions.
• Conclusions and recommendations — a six-part framework for ethical AI use: understand, verify, protect, supervise, communicate, and be truthful.
• Appendix — state-specific AI guidance.