-
Answer Quality Is Not Learning Impact: The Stanford AI-Tutoring Study and Hybrid Legal Education
Sixteen law professors judged nearly 3,000 blind comparisons between AI-generated and human-written answers to contracts questions, and preferred the AI 75 percent of the time. The finding is hard to dismiss and easy to overread. Hybrid JD programs that deliver most coursework asynchronously might see the study as a case for adopting AI tutors as the primary support mechanism for their distance components. The case is strong as far as it goes. The study measured whether professors preferred the AI’s answers, not whether those answers produce the learning outcomes that legal education is supposed to serve.
Read more -
Offensive Prompt Injection as Professional Misconduct
A Brazilian labor court has fined two lawyers for hiding white-text commands in a petition to manipulate the court’s AI, and the Pará bar has suspended them. U.S. bar opinions still treat the lawyer as the user of AI tools and the potential victim when those tools fail. The harder scenario has arrived: the lawyer embedding hidden instructions in a document to manipulate an adversary’s AI workflow. The Model Rules already prohibit the conduct. The bar should say so before a U.S. sanctions order has to.
Read more -
Reading the Limitations Section: What the First RCT on AI and Legal Reasoning Actually Shows
A new RCT finds that law students who used AI on an early synthesis task outperformed their peers on a later reasoning task, even without AI. The finding is encouraging, but conditional. The experimental design gave participants curated sources, decomposed tasks, structured prompts, and a selected model—the same institutional scaffolding most firms have not built. The study’s limitations section is a blueprint for the environment that produced the positive result.
Read more -
How Lawyers Should Prompt in 2026
Anthropic, OpenAI, and Google all updated their prompt engineering guidance in late 2025 and early 2026, and a striking amount of the advice from 2023 and 2024 now degrades the newer models. This post lists the changes practicing lawyers should know about, with concrete prompt patterns drawn from the three vendors’ current documentation.
Read more -
Client-Side AI Recording and NYC Bar Formal Opinion 2025-6
The NYC Bar’s Formal Opinion 2025-6 addresses what happens when clients use their own AI tools to record and transcribe conversations with their lawyers. Read alongside Heppner, the opinion establishes one clear duty (warning clients of AI-related privilege risks) and suggests two further responses (providing privilege-preserving alternatives and redesigning communication channels) that the rules do not yet require but that the foreseeability shift Heppner introduces makes worth taking seriously.
Read more -
SB 26-189's Obsolete Exemption: Statutory Categories and Moving Technology
Colorado’s SB 26-189 exempts AI tools used solely to ‘summarize, organize, translate, draft, route, or present information for human review.’ That exemption was drafted for a model of AI use—ask a question, get an answer, review the answer—that the legal technology market has already moved past. The tools law firms are buying don’t summarize information for lawyers. They make the analytical choices that determine what gets summarized, in what order, and with what emphasis.
Read more -
Open-Source Legal AI and the Institutional Floor
A former Latham associate reproduced the core features of Harvey and Legora in two weeks and released the code for free. The commentary has focused on what that means for pricing and capability. But Mike is a ceiling story, and the profession’s AI problem is at the floor.
Read more -
Agentic AI and the Boundaries of Professional Judgment
Legal technology vendors are marketing AI ‘agents’ that plan, reason, and execute multi-step workflows. These tools can handle information-gathering tasks well, including within legal practice itself. But the line between collecting material for a lawyer’s evaluation and substituting for that evaluation is the line between appropriate delegation and a supervisory problem under Rule 5.1—and the vendors’ incentives push firms to cross it.
Read more -
The Disclosure Patchwork
Illinois says courts should not require lawyers to disclose AI use. Florida circuits mandate it on the face of every filing. Hundreds of federal judges have issued individual standing orders, no two identical. The profession has spent three years arguing about whether disclosure is necessary without asking what disclosure is for—and the answer has less to do with catching errors than with enabling the people who review AI-assisted work to do their jobs.
Read more -
Citation Sanctions in Q1 2026: The Verification Problem, Quantified
U.S. courts imposed at least $145,000 in sanctions for AI-generated citation errors during Q1 2026 alone, across cases in New York, Kansas, the Sixth Circuit, and Oregon. The sanctioned lawyers share a striking common feature: none of them had functioning AI verification practices in place. That finding complicates the profession’s preferred response to AI risk, but not in the way you might expect.
Read more -
Revised Standard 314: Learning-Outcomes Requirements and the August Deadline
The ABA’s revised accreditation standards require law schools to establish measurable learning outcomes for every course, align them to programmatic outcomes, and build formative assessments into the first year—all by the 2026-2027 academic year. Most schools are not staffed for this work. An LLM can help with the drafting. It cannot help with the judgment calls that make the drafting worthwhile.
Read more -
Deployer Obligations Under the Colorado AI Act
Colorado’s AI Act takes effect on June 30, and its deployer obligations apply to anyone who uses AI as a substantial factor in consequential decisions—including law firms. ‘Legal services’ is one of the statute’s eight enumerated categories. Most of the legal profession has not grappled with the fact that it is on the regulated side of this law.
Read more -
Anatomy of an AI-Contaminated Filing: The Sullivan & Cromwell Errata
Sullivan & Cromwell’s AI-contaminated bankruptcy filing has drawn coverage for the firm’s apology. The three-page errata is more revealing: errors that suggest AI corrupted correct citations during editing, a compliance program that failed despite being rigorous, and a supervision obligation the firm’s letter concedes without naming.
Read more -
Upsolve and the Unauthorized-Practice Implications for Legal Chatbots
A federal court dismissed Upsolve’s challenge to New York’s unauthorized-practice-of-law rules, holding that trained non-lawyers cannot give individualized legal advice—even for free, even with safeguards, even with disclaimers. The opinion never mentions AI. But it describes AI legal tools more precisely than any opinion that has.
Read more -
New York S7263 and Chatbot Liability for Substantive Professional Advice
New York Senate Bill S7263 would impose civil liability on chatbot proprietors whose systems provide ‘substantive’ responses in areas reserved for licensed professionals—and declares that disclosing the chatbot’s non-human status is not a defense. The bill’s impulse is understandable, but its mechanism confuses information with advice and would suppress exactly the kind of public legal education that existing law permits.
Read more -
Building Infrastructure with AI: A Case Study
A law professor with no engineering background used Claude, Cowork, ChatGPT, and Gemini to design and deploy a self-hosted news aggregation pipeline over a weekend. The project worked—not because AI eliminated the need for technical skill, but because the skills it required turned out to be the ones lawyers already have.
Read more -
Sycophancy as a Failure Mode in AI-Assisted Legal Reasoning
Hallucination gets the headlines, but sycophancy may be the more dangerous failure mode for lawyers. An LLM that systematically validates your reasoning instead of challenging it functions as a mirror, not a counsel. And mirrors make poor advisors.
Read more -
A Delegation Framework for AI-Assisted Legal Work: Delegating the Task, Not the Judgment
LLMs are good at generating options, structuring information, and doing legwork. They are not good at weighing what they generate. The most common mistake lawyers make with AI is asking it to exercise judgment lawyers should be exercising themselves.
Read more -
Context-Window Degradation as a Practice Risk
LLMs degrade predictably as their context windows fill—losing track of middle-document content, dropping earlier conversation history, and producing confident output built on incomplete inputs. For lawyers using these tools on long documents, the question is not whether it happens but how to structure your work to prevent it.
Read more -
The Duty to Inform: Client AI Use as a Known Hazard After Heppner
Heppner established that consumer AI conversations are not privileged. But the case also raises an uncomfortable question for practicing lawyers: if a known hazard to the privilege now exists, do you have a duty to warn your clients about it? The answer, under existing ethics rules, is almost certainly yes.
Read more -
API Access and the Limits of Data-Handling Compliance
Using an LLM through an API rather than a consumer chatbot improves your data-handling posture—sometimes dramatically. But an API alone does not satisfy FERPA, HIPAA, or any other regulatory framework, and treating it as though it does mistakes a technical control for a legal one.
Read more -
Privilege and the Consumer Chatbot: Data Handling Across Claude's Tiers After Heppner
A comparison of Anthropic’s data-handling policies across Claude’s consumer and commercial tiers—and why the distinction now carries legal consequences after the SDNY’s decision in United States v. Heppner.
Read more