Ethics built in. Not bolted on.

What DDEAS is

DDEAS — Discover Docket Ethical and Accountability Standards — is the cryptographic backbone behind every output JILL produces. It is the framework that takes generative AI from "useful but unverifiable" to "useful and defensible."

The framework operates as four interlocking layers. Each layer corresponds to a specific failure mode of unsupervised AI in legal practice. Together, they constitute a closed verification environment that the AI cannot operate outside of.

Why this matters now

Courts have begun sanctioning lawyers for AI-fabricated citations. In Mata v. Avianca (S.D.N.Y. 2023), counsel submitted a brief citing six federal cases that did not exist. In Park v. Kim (2d Cir. 2024), a court referred counsel for discipline after an AI tool produced a non-existent authority. The list continues to grow.

Guidance has followed the cases. ABA Formal Opinion 512 (2024) maps a lawyer's existing duties onto generative-AI tools, and a growing number of state bars have issued their own. The throughline is the same: the duty of competence and candor does not transfer to the tool. DDEAS is built so that duty is supportable.

The four layers

Layer 1 — Source validation. Before any citation JILL produces appears on your screen, the citation is checked against the authoritative case databases for the jurisdiction your matter sits in. If the case doesn't exist, the citation doesn't appear. If the citation is slightly malformed — a misnumbered volume, a transposed reporter, a misattributed circuit — the citation is corrected or surfaced as requiring review. The hallucinated case never reaches you because the validation layer is the gatekeeper between the model and your work product.

This is the structural difference between Discover Docket and any other AI legal tool. Generic AI validates nothing. JILL cannot return a citation that hasn't cleared validation.

Layer 2 — Confidence scoring. Every legal proposition, every citation, every fact assertion JILL produces carries an explicit confidence score from 0 to 100. A score at or above 95 means the conclusion was drawn from validated, current authority with strong textual support. A lower score signals that you should verify more deeply before relying on it.

Confidence scores are produced by independent scoring models that evaluate each output against the source material, the consistency of supporting authority, and the freshness of the underlying data. They appear next to every paragraph of a drafted motion, every cited authority, every analytical conclusion. You always know where the certainty is — and where it isn't.

Layer 3 — Cryptographic audit log. Every action JILL takes — every prompt you give her, every authority she retrieves, every output she returns, every score she assigns — is recorded in an append-only audit log. Each entry is timestamped to the millisecond, signed with a cryptographic key tied to your firm, and chained into the previous entry's hash so the entire log forms a tamper-evident sequence.

The log is filterable by matter, by date range, by user, by output type. It serves two audiences: the supervising attorney who needs to review associate work, and the court that may someday need to see exactly how a brief was researched and drafted.

Layer 4 — Tamper-evident chain. The audit log uses the same hash-chain architecture that secures financial transaction systems and digital evidence in criminal cases. Each entry's hash includes the previous entry's hash. Altering any historical record breaks the chain, which is detectable on inspection.

This is what makes the audit log evidence rather than documentation. A spreadsheet of your AI usage can be edited after the fact and is therefore worthless as defensibility material. The DDEAS audit log can be authenticated.

What "defensible" actually means

The word "defensible" is doing real work here. In the post-Mata v. Avianca, post-ABA Opinion 512 environment, every lawyer using AI tools needs to be able to answer a specific question if challenged: how do you know your work product wasn't AI-fabricated?

The answers fall on a spectrum.

A lawyer using ChatGPT carefully, with manual verification, who keeps no contemporaneous record of which authorities were checked or when — that lawyer's answer is "I remember verifying it." If opposing counsel files a sanctions motion alleging the brief contains AI fabrications, the lawyer's defense is testimony. Testimony is the weakest form of evidence the legal system recognizes.

A lawyer using ChatGPT who manually maintains a spreadsheet of "AI sessions" with timestamps and outputs — that lawyer's answer is "here is my log." It's a better answer. But a spreadsheet can be edited. Opposing counsel can credibly argue the log was reconstructed after the fact.

A lawyer using a platform with cryptographically signed and chained audit logs can produce a record that is mathematically tamper-evident. The hash chain either authenticates or it doesn't. There is no in-between. That lawyer's answer to the challenge is the same shape as the answer in any chain-of-custody dispute: here is the contemporaneous record, signed, time-stamped, unalterable, and produced from the system at the time the work was done.

That is what defensibility means. It is not "I followed best practices." It is "I can prove I followed best practices, and the proof is not my testimony, it is the system."

Founder POV — why we built it this way

I've been in the legal industry for twenty five years. Long enough to remember when the duty of competence under Rule 1.1 meant knowing the rules of evidence and being able to find a case in a bound reporter. Long enough to have watched competence expand to include electronic discovery, then social media evidence, then encrypted communications, and now generative AI.

The expansion of competence is the expansion of risk. Every new layer of technology a lawyer is required to understand is a new layer at which the lawyer can fail. The duty does not pause because the technology is hard. The duty grows.

Generative AI is the largest expansion of the competence duty in my professional lifetime. It is also the one that most lawyers are least equipped to handle, because the failure modes of generative AI — confidently asserting false information, fabricating citations, mimicking authority that does not exist — are precisely the failure modes that lawyers are professionally trained to take at face value when produced by a credible-looking source. The model looks credible. It produces output that looks credible. The lawyer who has not been schooled in how generative AI fails will, by default, treat it as a credible source.

The legal profession will not move backwards. Generative AI is going to be part of practice for the rest of my career and well beyond it. The question is not whether lawyers will use it. The question is whether the lawyers using it will be the ones who can defend the work afterward.

I built Discover Docket — and inside it, DDEAS — because I want to be one of those legal professionals, and I want to give every solo, every small firm, and every mid-sized practice in the country the infrastructure to be one too. Practice with the technology. Defend the work. Sleep at night.

— Chris Waters

Privacy commitment

DDEAS protects every AI action inside the platform. Discover Docket also operates without cookies or tracking software on the marketing site, and does not sell or share email addresses or contact information.