Where the AI Stops and the Lawyer Starts

The platform is an aid to legal work, not legal advice. It flags, surfaces, and recommends. It never forms an attorney-client relationship, gives legal advice, or makes a legal decision. A licensed human decides.

This page draws the line the whole system is built around: everything left of it is software's job, everything right of it is a lawyer's. The boundary is not a policy disclaimer. It is the operating contract between the code and the profession it serves.

The UPL Line

The System

●Classifies and routes documents to the right matter

●Extracts terms, obligations, and parties

●Scores risk and attaches a confidence level

●Flags problem language and suggests redlines

●Surfaces relevant precedent and firm standards

●Escalates low-confidence output to a human

The UPL Line

The Attorney

●Forms the legal conclusion and the legal judgment

●Gives legal advice to the client

●Chooses the strategy and the negotiating position

●Signs, files, and accepts responsibility for the work

●Communicates directly with the client

Everything the system produces is input to a licensed decision, never the decision itself. The red line is crossed the moment software stands in for the lawyer instead of preparing the lawyer's work.

The Rules It Cites

ABA Model Rule 5.5

Unauthorized Practice of Law; Multijurisdictional Practice

The rule the boundary exists to respect. It confines the practice of law to licensed lawyers and makes clear that legal services are the lawyer's province. The platform is designed so that everything it does stays outside the zone Rule 5.5 protects.

ABA Model Rule 5.3

Responsibilities Regarding Nonlawyer Assistance

The ethical frame for AI as nonlawyer assistance under lawyer supervision. The Comment requires the lawyer to make reasonable efforts to ensure that a nonlawyer's conduct is compatible with the lawyer's professional obligations. The system is the nonlawyer assistant, and it is built to stay inside that supervised lane.

ABA Model Rule 1.1, Comment 8

Duty of Technological Competence

A lawyer must keep abreast of the benefits and risks associated with relevant technology. Using the platform is competent practice only because it is transparent, verifiable, and supervised. Chain-of-reasoning and source attribution are what make the tool a competence enabler rather than a competence risk.

ABA Model Rule 1.6

Confidentiality of Information

Client information is protected against disclosure. The platform enforces this with ethical walls and data isolation: practice groups are walled, client data stays in its own tenant, and nothing leaves the trust boundary for third-party training. Confidentiality is structural, not a checkbox.

ABA Formal Opinion 512 (2024)

Generative Artificial Intelligence Tools

Formal Opinion 512 holds that a lawyer may use generative AI while discharging their duties of competence, confidentiality, and supervision, so long as the lawyer maintains ultimate responsibility for the work. The platform's design is consistent with it: the lawyer verifies every output, the data stays protected, and the model is supervised rather than autonomous.

Design Guarantees That Respect the Line

Confidence Thresholds Escalate

Low-confidence output is never released. Below the threshold, results are pushed to a human reviewer before anything reaches a matter. Risk flags escalate regardless of confidence.

Rule 5.3 · 5.5

No Autonomous Legal Decisions

The system cannot form a legal conclusion, recommend a specific legal outcome, or take a legal action on its own. Autonomy is capped at preparation and surface. The decision always lives with the attorney.

Rule 5.5

Human-in-the-Loop for Every Decision

Every classification, risk score, and redline is reviewed before it becomes part of a lawyer's work product. Every human override is logged with who, when, and why. No output reaches the client unexamined.

Rule 5.3 · Op. 512

Chain-of-Reasoning, Verifiable

Every recommendation carries its reasoning, visible to the reviewer. The attorney can trace how the system reached an answer, question it, and override it. The tool is transparent by construction, not by policy.

Rule 1.1 Cmt 8

Source Attribution

Every conclusion cites the clause, standard, or precedent that drove it. Nothing is a black box. If an attorney questions a result later, the full context is retrievable and attributable.

Rule 1.1 Cmt 8

Ethical Walls and Data Isolation

Practice groups are walled from each other. Client data stays in its own tenant and never leaves the trust boundary. Confidentiality is enforced at the data layer, mapping directly to the lawyer's duty under Rule 1.6.

Rule 1.6

What This Means for an Interviewer

A non-lawyer legal engineer has to know exactly where the role stops. This page is that map. The honest answer to "who's responsible for the legal work" is: the licensed attorney, every time. The engineer's job is to build a system the lawyer can trust and verify, not to build a lawyer. When an interviewer asks where the AI ends and the profession begins, the answer is a drawn line, not a hand wave. Build the side of the line that prepares the work, and make it transparent enough that a lawyer is confident standing behind it.