What Cubby actually launched
Cubby announced on September 3 that it rebuilt its student platform and introduced Cubby for Professors, a faculty-facing tool that lets law school instructors deploy AI practice and feedback tools calibrated to their own rubrics and teaching style rather than a generic model's default behavior. More than 20 law schools are running the professor tool for fall 2026, including Emory, Wake Forest, the University of Texas, Case Western Reserve, and Quinnipiac, a list that spans elite private programs and large public law schools alike, suggesting the pitch resonates regardless of an institution's size or prestige tier.
Founder and CEO Truman Sacks framed the underlying problem in plain terms in the company's announcement: professors want to give students more practice and feedback, but there are not enough hours in a week to deliver it one-on-one at scale. Cubby's bet is that faculty will hand a narrow, well-defined slice of that workload to AI, provided they retain control over exactly what the AI says, how it grades, and which materials it can see, a bet that this week's expansion suggests is paying off with faculty who were previously skeptical of any AI tool touching graded work.
The scale behind the announcement
Cubby says it served nearly 2,000 students across more than 200 U.S. law schools in its first year, a base it built before this week's professor-facing launch and largely through individual student adoption rather than institutional contracts. That is a meaningful footprint in a market where law school administrations have been openly hostile to general AI chatbots over concerns about hallucinated case citations, degraded legal writing skills, and undermined bar exam preparation across an entire graduating class of aspiring attorneys who will eventually need to pass a licensing exam without an AI assistant in the room.
The expansion to more than 20 schools running the professor tool specifically is the more important number for anyone tracking institutional adoption rather than consumer downloads. It signals a shift toward formal procurement rather than individual student subscriptions, a harder sale in legal education generally, and a stronger indicator that the product is solving a problem law school administrations recognize as legitimate rather than one only anxious students care about on their own.
Governance as the actual product
Emory Law associate dean Kamina Pinder's endorsement, that Cubby made her a believer that AI can enhance rather than replace critical thinking, is the kind of quote every AI-education vendor chases for its marketing materials. The substance underneath that quote is procedural rather than technical: Cubby does not share one professor's materials with another institution, does not use faculty content to train its underlying model, and lets instructors exclude past exams from what students can access through the platform, terms that a skeptical associate dean would have needed to see in writing before attaching her name to any public endorsement.
Those three guarantees function as the actual differentiator in this deal, more than the AI capability itself. Any competent large language model can generate practice questions or draft feedback on a legal memo. What law schools could not get from ChatGPT or a generic AI grading tool was a contractual, technical promise that a professor's proprietary exam bank would not leak into training data or surface, even indirectly, inside a rival school's version of the same product, a guarantee general-purpose consumer chatbots are not built to make to any single institutional customer.
Why law schools are the proving ground
Legal education sits among the highest-stakes assessment environments in all of higher education, where a single exam often determines a full semester's grade and eventually feeds into readiness for a licensing exam that gatekeeps an entire profession. That combination makes law schools an unusually strict test of whether an AI tool can be trusted anywhere near grading-adjacent work, which is exactly why faculty buy-in here required far more than a technically capable underlying model.
It also makes law schools a useful proxy for other high-stakes enterprise training environments, from financial services compliance certification to medical licensing prep, where a nearly identical objection applies. No enterprise buyer in a regulated industry wants a vendor's model quietly absorbing proprietary assessment content and later redistributing patterns of it to a competitor, a regulator's audit, or another customer running the same platform under a different contract, particularly in industries where certification exams themselves are tightly controlled intellectual property.
The lesson for enterprise buyers
Corporate learning teams shopping for AI-enabled coaching or assessment tools should read Cubby's governance language as close to a template contract clause worth copying wholesale into their own procurement documents. Data segregation between clients, an explicit no-training-on-our-content guarantee, and the contractual ability to exclude sensitive assessment material from the AI's accessible corpus are specific, auditable commitments, not vague trust-us language buried in a terms-of-service page nobody reads before signing, and each one can be verified through a standard vendor security review rather than taken on faith.
This is also a market signal worth taking seriously on its own terms: buyers in a sector this anxious about AI accuracy and intellectual property leakage chose a narrow, governed tool over a general-purpose one available for free. Enterprise CTOs weighing build-versus-buy decisions on AI coaching tools should expect that same preference to hold in corporate compliance and skills training, where proprietary assessment content carries the same sensitivity as a law professor's carefully guarded exam bank, and where a leaked certification question can trigger a regulatory finding rather than just an academic integrity complaint.
