The sanctions orders are public. The carriers have added AI questions to renewal applications. Clients' general counsel are writing AI clauses into outside-counsel guidelines. What a firm needs is not another ethics CLE; it is evidence that supervision, confidentiality and review actually happen. AboveBoard verifies that evidence and gives you a score to show for it.
The largest hallucination sanction on record is roughly $109,700 in Couvrette v. Wisnovsky (D. Or., Dec 2025); Whiting v. City of Athens (6th Cir., 2026) imposed $15,000 per attorney plus fees and double costs; more than $145,000 in penalties were ordered in Q1 2026 alone (haqq.ai summary of the Charlotin database, Jun 2026). Many federal and state judges now have standing orders requiring disclosure or certification of generative-AI use. Where a tool was identified, ChatGPT accounted for 53%, Westlaw 18%, Claude 15% and Copilot 12% (Stanford CIS, Oct 2025).
ABA Formal Opinion 512 (Jul 29, 2024) maps generative AI onto competence, confidentiality, communication, fees and supervision. By August 2026 a 50-state tracker counted 38 jurisdictions with official AI guidance, 17 of them formal ethics opinions, including DC, Florida, Texas, New York City, North Carolina, Pennsylvania and Virginia (legalaicompliance.help, Aug 2026). Utah's AI Policy Act requires licensed professionals to disclose generative AI in high-risk interactions (as amended 2025). Colorado's replacement law, SB 26-189, reaches decisions about legal services from January 1, 2027.
EPIC's May 2026 survey of thirteen carriers put it plainly: "AI-related malpractice exposure has moved from theoretical to real." Industry compilations report CNA adding supplemental AI questionnaires at renewal, at least one carrier excluding named generative-AI tools, and underwriters asking "Do you use AI? Do you police it? Do you have protocols in place?" (legalaigovernance.com and actuary.info, 2026; secondary compilations). ISO introduced a generative-AI exclusion for general liability policies in January 2026 (Fenwick, Jun 2026).
More than half of corporate legal departments want their outside firms to use AI, yet 68% have no idea whether they do, and 40% of firms get conflicting client instructions on AI (Thomson Reuters, Feb and Mar 2026). Hence the AI sections now in outside-counsel guidelines and RFPs (ABA Law Practice Magazine, Jul/Aug 2026): which tools, on whose data, with what review, and can you prove it.
AI assistants inside Westlaw, Lexis and practice-management platforms; dedicated legal models such as Harvey and CoCounsel; contract analysis and clause libraries. Best terms, still capable of fabricated authority.
ChatGPT, Claude, Gemini and Microsoft 365 Copilot for summaries, first drafts, correspondence and deposition prep. The category most often used on personal accounts and least often disclosed. Adoption runs 71% at solo firms, 75% at small firms and 86% at mid-sized firms (Clio Legal Trends via NC Bar, May 2026).
E-discovery and technology-assisted review, intake chatbots on the website, meeting and call transcription, time-entry narrative drafting, billing review. Each touches privileged or confidential material and belongs in the inventory.
Forty controls, eight per domain, scored against evidence. Here is how they read when the firm's obligations come from the Rules of Professional Conduct.
A written AI policy adopted by the partnership, a named partner or general counsel with authority to halt a tool, an inventory of AI tools by practice group, and an obligations register that tracks ethics opinions and standing orders in every jurisdiction where you appear.
Evidence examples: the policy with the executive committee's approval date; the register listing research AI, drafting AI and intake chatbots with an owner for each; the obligations memo summarizing Opinion 512 and your state bar's guidance, updated within six months.
Confidentiality under Rule 1.6 and privilege translated into a classification staff can apply in seconds, an approved-tool list with enterprise terms that prohibit training on inputs, single sign-on, retention settings for prompts and outputs, and a consent path for matters where a client must agree first.
Evidence examples: the vendor terms review log with the no-training and sub-processor clauses highlighted for each approved tool; SSO and offboarding configuration; the informed-consent template used before entering confidential information into any self-learning tool, as Opinion 512 contemplates.
A written rule that every citation and factual assertion in AI-assisted work is verified by a named lawyer before filing or sending, recorded spot-checks, an incident plan that covers a hallucinated citation reaching a court, and matter-level tagging of AI-assisted work product.
Evidence examples: the pre-filing cite-check procedure and a sample sign-off; the incident playbook step for notifying a client and correcting the record with the court; the document-management field that records which matters used AI and for what.
Training completed by 90%+ of lawyers and staff, role-specific modules, reviewer training on verifying AI output, and signed acknowledgments. The 8am 2026 survey found 54% of legal professionals had received no responsible-AI training (NC Bar, May 2026); this is where firms gain points fastest.
Evidence examples: the completion report with dates and roles; the acknowledgment file; the partner-level governance training record.
A standard AI clause in engagement letters, a published AI statement, a procedure for court and regulator disclosure where required, a review step for AI claims in pitches and RFP responses, and a standard answer bank for outside-counsel guideline questionnaires.
Evidence examples: the engagement-letter paragraph and the opt-out handling; the court-disclosure procedure keyed to standing orders; the last three OCG questionnaire responses, consistent with each other and with the evidence file.
No firm is listed at any level with a zero on G1 (written policy), D3 (vendor terms reviewed) or O1 (human review before client-facing output). For a law firm, O1 is the control that separates a sanctions order from a near miss.
Read the full rubricAnswered once, from verified evidence, with a registry link for the client's general counsel. Inconsistent answers across pitches lose panel spots.
The application's AI questions map to G1, D2, D3, O1, O3 and W1. Attach the report and the year-stamped badge; coverage decisions remain the carrier's.
A board-ready report by domain, with each gap and its fix, that a managing partner can present in fifteen minutes. Improvement shows in the registry at renewal.
AboveBoardAI issues a score and a verification under a private, voluntary standard; it is not a bar certification, not legal advice, and not a guarantee that any filing is accurate. It verifies that the practices which prevent bad filings exist and are evidenced. See What is an AI audit? and our impartiality rules.
Take the free Snapshot. Ask each practice-group leader which AI tools their lawyers and staff used in the last 90 days, including personal accounts. Write it down; that is your register.
Adopt an AI acceptable-use policy (start from our template). Approve specific tools on firm accounts with enterprise terms. Add one sentence to the filing procedure: no AI-assisted citation goes out unverified by a named lawyer.
Roll out Academy Foundations to every lawyer and staff member; collect signed acknowledgments. Brief the executive committee for thirty minutes on their oversight responsibilities and minute it.
Insert the AI clause into new engagement letters and update the OCG answer bank. Start the full assessment; Snapshot answers carry over.
Growth Firm (50–249) $18,000; Established Firm (250–999) $42,000; Enterprise (1,000+) from $75,000. Every tier includes the assessment, independent review, score, badge, registry listing, board report and Academy Foundations. ISO/IEC 42001, by comparison, runs about $73,000 in year one for a 30-person firm (certbetter, Jun 2026).
Opinion 512 (July 29, 2024) ties generative AI to the existing duties of competence, confidentiality, communication, reasonable fees and supervision. Disclosure depends on the circumstances: informed consent is expected before confidential client information is entered into a tool that may learn from it, and clients should be told when AI use affects fees or the representation. Many firms handle this with a standard engagement-letter clause plus case-by-case consent for sensitive matters.
The Charlotin AI Hallucination Cases database counted 1,598 court cases worldwide involving AI-fabricated content as of June 2026, growing by roughly eight a day. A Stanford CIS analysis (October 2025) found that about 90% of lawyer-filed AI-tainted filings in the US came from solo practitioners and firms of 2 to 25 lawyers.
Each carrier sets its own underwriting requirements, and AboveBoard makes no promise about coverage or premium. What the report gives you is a dated, independently reviewed answer to the questions carriers are now asking: whether you have a written AI policy, an approved-tool list, a human-review rule, training records and an incident plan. Seven of thirteen lawyers' professional liability carriers surveyed by EPIC in May 2026 now report AI-related claims.
Yes. Research-platform AI has better contractual terms than a consumer chatbot, but it can still produce fabricated or mis-cited authority, and the Stanford CIS data attributes 18% of identified AI-tainted filings to Westlaw tools. A policy also covers the tools you do not know about: Clio found that 57% of solo and 55% of small firms have no AI policy even though 71 to 75% have adopted AI (NC Bar, May 2026).
AboveBoardAI is priced annually by headcount: $7,500 for firms of 1 to 49 people, $18,000 for 50 to 249, $42,000 for 250 to 999 and from $75,000 for 1,000 or more. Every tier includes the assessment, independent review, score, badge, registry listing, board report and Academy training, which for lawyers is built to align with CLE ethics content. See pricing.
No. AboveBoardAI issues a score and a verification under a private, voluntary standard. No bar association, court or regulator endorses it. It is designed so the evidence it verifies is the same evidence a bar counsel, a client's general counsel or a carrier would ask to see.