AI Professional Ethics

State of play.

• Sanctions for AI hallucinations have moved from isolated incidents to a documented enforcement pattern across jurisdictions. Tennessee ($45K, disciplinary referral), Connecticut (CLE mandate, $3,500 fine), and Delaware (show-cause against Richards Layton in Leiske v. Kidd) represent distinct enforcement vectors — federal Rule 11, state supreme court discipline, and Court of Chancery Rule 11(c) — all converging on the same conduct (→ Tennessee Firm Sanctioned $45K for AI-Generated Fake Citations in Malpractice Suit, Connecticut Supreme Court orders lawyer CLE after ChatGPT citation errors, Connecticut Judge Fines Lawyer $3,500 Over AI-Fabricated Quotes in Taco Trademark Case, Delaware court weighs sanctions after Richards Layton’s AI-generated brief errors).
• Shadow AI has emerged as the dominant in-house governance failure. A survey cited in Law360 found 83% of legal teams use AI tools not provided by their employer, 81% use unapproved tools, and 47% operate without any AI-use policy — creating unmapped exposure across privilege, confidentiality, data protection, and malpractice .
• The "negligence for non-use" argument has entered the professional discourse. A University of Houston law professor has publicly raised whether failure to use AI could itself constitute a breach of reasonable diligence — a direct inversion of the current sanctions regime — signaling that the profession is approaching an inflection point on AI as a competence floor, not just a liability risk (→ Lawyer Faces Potential Negligence Claim for Failing to Use AI in Legal Work).
• Major firms are formalizing AI governance structures. Reed Smith has established a dedicated AI department with over 700 lawyers on Harvey AI daily and firm-wide associate certification, signaling that AI integration is shifting from competitive differentiator to baseline infrastructure (→ Reed Smith Launches AI-Focused Legal Services Team Led by Philadelphia Partner).
• For counsel advising law firms or managing litigation teams, the practical baseline is that independent citation verification is now a nondelegable duty courts will enforce through sanctions regardless of remedial measures taken after the fact, and shadow AI in the client's legal department creates privilege and confidentiality exposure that outside counsel cannot ignore.

Where things stand.

• ABA Formal Opinion 512 remains the national ethics baseline but courts are imposing stricter standards through sanctions. The opinion's permission to reduce verification for "familiar" tools is internally inconsistent with hallucination data the opinion itself cites, and the sanctions pattern demonstrates that courts are not waiting for bar guidance .
• State bars are diverging. California's COPRAC has proposed amendments to six Rules of Professional Conduct imposing binding verification obligations with no routine-matter exceptions; Mississippi adopted Opinion 512 verbatim without independent analysis — the gap between the most and least demanding state standards is widening .
• Fake-citation sanctions are now a multi-jurisdictional enforcement record. Tennessee (Reaves Law, $45K, disciplinary referral), Connecticut (Gottlieb, CLE; Miller, $3,500), Delaware (Richards Layton, show-cause pending), California ($31K sanctions), Georgia (Supreme Court overturned AI-hallucinated order), and the 7th Circuit (former immigration judge admonished) — each adds to the record courts cite when imposing penalties (→ Tennessee Firm Sanctioned $45K for AI-Generated Fake Citations in Malpractice Suit, Connecticut Supreme Court orders lawyer CLE after ChatGPT citation errors, Connecticut Judge Fines Lawyer $3,500 Over AI-Fabricated Quotes in Taco Trademark Case, Delaware court weighs sanctions after Richards Layton’s AI-generated brief errors, Lawyer Faces Potential Negligence Claim for Failing to Use AI in Legal Work).
• Connecticut has adopted a Practice Book rule requiring independent verification of all AI-generated citations, authorities, and evidence before submission — the first state to codify the obligation in court rules rather than ethics opinions alone, with sanctions expressly available for violations (→ Connecticut Supreme Court orders lawyer CLE after ChatGPT citation errors, Connecticut Judge Fines Lawyer $3,500 Over AI-Fabricated Quotes in Taco Trademark Case).
• The privilege framework for AI-generated materials remains unsettled. SDNY's Heppner held documents generated on a public AI platform are not privileged absent attorney direction; E.D. Michigan's Warner v. Gilbarco found privilege intact where attorneys directed AI use — the split turns on attorney direction and platform confidentiality terms, neither of which is settled doctrine .
• Shadow AI in legal departments is a systemic governance failure, not an edge case. The documented exposure vectors include confidentiality breaches, privilege waiver, data protection violations, malpractice liability, and regulatory non-compliance — and most legal departments cannot currently account for where sensitive work is being processed .
• Agentic AI governance is the next compliance frontier. The "human-at-the-helm" framework — pre-deployment parameter-setting and tiered risk controls rather than post-hoc output review — is the emerging standard, but no bar or court has specified what those controls must look like in practice .
• Discovery workflows face a judicially imposed human-judgment floor. White v. Walmart (S.D. Ind.) established that AI cannot satisfy the attorney's independent obligation to review, narrow disputes, and meet and confer in good faith .
• The DOJ has established an AI Litigation Task Force, formalizing AI's role in federal legal operations and signaling that institutional AI governance is now a baseline expectation for sophisticated legal practice (→ DOJ Establishes AI Litigation Task Force as Courts Adapt AI Discovery Tools).
• AI accelerates discrete litigation tasks but does not compress case timelines. Every AI output still requires human approval before filing; the practical effect is workflow acceleration for document processing, not a reduction in attorney judgment requirements (→ DOJ Establishes AI Litigation Task Force as Courts Adapt AI Discovery Tools, Above the Law Says Legal Judgment Still Can’t Be Automated).

Latest developments.

• Shadow AI survey documents that 83% of legal teams use employer-unapproved AI tools and 47% operate without any AI-use policy, framing unauthorized AI deployment as a systemic in-house governance crisis .
• Above the Law commentary argues AI can automate contract summarization and clause drafting but cannot replicate business-context judgment on which issues warrant negotiation — framing human judgment as the durable competitive advantage in in-house practice (→ Above the Law Says Legal Judgment Still Can’t Be Automated).
• University of Houston professor Renee Knake Jefferson raises the novel argument that failure to use AI could constitute negligence, inverting the current sanctions regime and signaling an emerging tension between AI-as-risk and AI-as-competence-floor (→ Lawyer Faces Potential Negligence Claim for Failing to Use AI in Legal Work).
• Richards, Layton & Finger faces a show-cause order in Delaware Court of Chancery under Rule 11(c) after a January 2026 brief in Leiske v. Kidd contained fabricated citations and invented quotations, with a subsequent correction leaving material inaccuracies unresolved — the highest-profile firm-level AI sanctions proceeding to date (→ Delaware court weighs sanctions after Richards Layton’s AI-generated brief errors).
• Connecticut federal judge imposes $3,500 sanction on attorney Hilary B. Miller for AI-generated fabricated quotations and false citations in a trademark dispute, finding "subjective bad faith" despite the attorney's stated remedial measures (→ Connecticut Judge Fines Lawyer $3,500 Over AI-Fabricated Quotes in Taco Trademark Case).
• Connecticut Supreme Court sanctions attorney Ian Gottlieb — the court's first AI-hallucination sanctions ruling — ordering six hours of CLE in ethics and law office management for ChatGPT citation errors in landlord-tenant appeals; Connecticut's new Practice Book verification rule is now in force (→ Connecticut Supreme Court orders lawyer CLE after ChatGPT citation errors).
• Tennessee federal court orders Reaves Law to pay $45K to Baker Donelson under Rule 11 for fabricated citations in a malpractice suit, refers the firm to the Tennessee Board of Professional Responsibility, and circulates the sanctions order to other district judges (→ Tennessee Firm Sanctioned $45K for AI-Generated Fake Citations in Malpractice Suit).
• Above the Law analysis challenges the assumption that AI reduces attorney hours, arguing that validation and oversight labor consumes most efficiency gains — with Thomson Reuters data suggesting theoretical savings of 190-240 hours annually that the article contends are largely offset by system management costs (→ Above the Law article argues AI-first law firms work smarter hours, not fewer, due to machine management demands).
• Reed Smith establishes a dedicated AI department combining lawyers, e-discovery specialists, and AI professionals, with over 700 lawyers on Harvey AI daily and firm-wide associate certification (→ Reed Smith Launches AI-Focused Legal Services Team Led by Philadelphia Partner).

Active questions and open splits.

• Does the ABA's "familiar tool" verification reduction survive scrutiny? Opinion 512's permission to reduce independent review for tools attorneys know well conflicts with the hallucination data the opinion itself cites — and Connecticut's new Practice Book rule and California's proposed binding rule both reject any such exception, creating a direct conflict between the ABA baseline and the emerging state standard (→ Connecticut Supreme Court orders lawyer CLE after ChatGPT citation errors).
• What firm-level governance satisfies Rule 11 and Rule 11(c) when AI-generated errors reach filed documents? The Richards Layton show-cause proceeding will test whether a major firm's internal review process and post-filing correction effort are sufficient — or whether courts will impose structural governance requirements as a condition of avoiding sanctions (→ Delaware court weighs sanctions after Richards Layton’s AI-generated brief errors).
• Is "subjective bad faith" the standard for AI citation errors even after remedial measures? The Connecticut federal court found bad faith despite Miller's apology and stated protocol revisions — if that framing holds, the intent inquiry in AI sanctions cases is more demanding than attorneys may assume (→ Connecticut Judge Fines Lawyer $3,500 Over AI-Fabricated Quotes in Taco Trademark Case).
• Will the Heppner/Warner privilege split be resolved at the appellate level? The two district courts reached opposite conclusions on nearly simultaneous facts; the determinative variables — attorney direction, platform confidentiality terms — are not yet settled doctrine, and clients using public AI platforms remain exposed until a circuit weighs in .
• When does shadow AI use by a client's employees create outside counsel exposure? If a client's legal team is processing privileged materials through unapproved consumer AI platforms without counsel's knowledge, the privilege waiver and confidentiality breach risks flow back to the engagement — but no court or bar has defined outside counsel's duty to audit or advise on client-side AI governance .
• Is failure to use AI becoming a cognizable negligence theory? The "negligence for non-use" argument has entered professional discourse but has no decided case behind it — the tension between courts sanctioning AI misuse and academics arguing AI non-use may breach diligence creates a narrowing window for practitioners to define their own standard (→ Lawyer Faces Potential Negligence Claim for Failing to Use AI in Legal Work, Above the Law Says Legal Judgment Still Can’t Be Automated).
• What governance architecture satisfies the duty of competence for agentic AI? The "human-at-the-helm" framework calls for pre-deployment controls and tiered risk management, but no bar or court has specified what those controls must look like — and the shadow AI problem demonstrates that governance frameworks that exist on paper are not being implemented in practice .

What to watch.

• Vice Chancellor Will's ruling in Leiske v. Kidd — whether sanctions are imposed on Richards Layton, Richard Rollo, or both, and whether the court articulates governance standards firms must meet to avoid joint liability under Rule 11(c).
• Whether the California Supreme Court adopts COPRAC's proposed amendments and whether the final rule retains the no-exceptions verification standard — the first binding state ethics rule on AI will set the national benchmark.
• Whether any circuit court takes up the Heppner/Warner privilege split — the first appellate ruling will define platform selection and client counseling obligations firm-wide.
• Whether the Tennessee Board of Professional Responsibility proceeding against Reaves Law produces a published disciplinary ruling that operationalizes what AI verification protocols must look like.
• Whether additional state bars or federal district courts adopt Connecticut-style mandatory verification rules, or whether courts continue imposing the standard through sanctions without codifying it.
• Whether the "negligence for non-use" argument produces a bar opinion or malpractice claim — the first decided case or formal ethics opinion on that theory will shift the competence calculus for practitioners who have not yet integrated AI into their workflows.