Legal & Regulatory

Before any model reasons, the companion works the question with you in plain conversation. Jurisdiction and circuit. Procedural posture. Standard of review. What is already decided and not up for relitigation. The client's actual tolerance for risk. You approve the brief. The deliberation starts from your matter, not from a generic reading of the question.

• ·CourtListener — federal and state opinions, SCOTUS, U.S. Tax Court, PACER/RECAP dockets
• ·govinfo (GPO) — the U.S. Code, public laws, the Congressional Record — resolved by citation
• ·eCFR — the Code of Federal Regulations as it reads today, part by part
• ·Federal Register — final rules, proposed rules, agency notices, presidential documents
• ·Congress.gov — bills, resolutions, amendments, committee reports
• ·Regulations.gov — rulemaking dockets and the public comment record behind a rule
• ·State statutes — official legislature text for 17 states and the District of Columbia, addressed by citation
• ·New York Open Legislation — the NY consolidated laws in full text, from the State Senate
• ·OpenStates — pending legislation across all 50 state legislatures
• ·SEC EDGAR — full-text search across public company filings
• ·OFAC SDN — Treasury's Specially Designated Nationals list
• ·EUR-Lex · HUDOC · legislation.gov.uk · Légifrance — for the matters that do not stop at the border

Statutes and regulations are fetched by citation, from the publisher's own text — hand the panel 15 U.S.C. § 78m(d) or 17 C.F.R. § 240.10b-5 and it reads the section, not a summary of the section.

Every claim carries a provenance tag. [SOURCED] means traced to a primary authority. [INFERRED] means analytical inference — flagged automatically, so the reasoning that rests on judgment is visibly separate from the reasoning that rests on authority.

Case citations are checked before you see them. The reporter cite, the court, the year and the party names have to match a real opinion in CourtListener, or the claim is flagged. That is a fabrication check, not a citator: it establishes that the case exists, not that it still controls.

• The Architect maps the structure — the chain of obligations, the deal mechanics, the operational facts a court would have to find.
• The Strategist reads the enforcement posture — what the agency has actually done, where the circuits diverge, how the timing runs.
• The Engineer tests the mechanics — thresholds, deadlines, notice periods, the arithmetic inside a damages or penalty calculation.
• The Counsel states the exposure plainly — liability, duty, privilege, and the conflicts between the jurisdictions in play.
• The Contrarian one mandate: argue the other side. Every authority, every inference, every fact assumed rather than established.

Five complete analyses, written blind so none can anchor on another. Then they cross-examine each other anonymously — labeled ANALYSIS_1 through ANALYSIS_5, so the argument is judged on the argument. Then they converge.

Output: GO / PIVOT / NO GO / INSUFFICIENT BASIS · Confidence score · Decision matrix · Minority Report · Testable predictions.

Questions Pilot5.ai helps with

• Does this board decision survive an entire-fairness challenge in the Delaware Court of Chancery?
• What is our real exposure under the CCPA and the state privacy statutes that followed it?
• The comment period on this proposed rule closes in thirty days. Do we file, and what does silence cost us?
• Are these non-competes enforceable in the states where the employees actually work?
• Where do the circuits diverge on this question, and which one are we sitting in?
• Our indemnity turns on “gross negligence.” Does that carry the same weight under New York law as under Delaware law?

Common questions

Is Pilot5.ai compatible with privilege and client confidentiality?

Your matters are never used to train any AI model. Before a deliberation reaches the models, a pseudonymization layer replaces personal and organization identifiers with reversible tokens, and the original terms are restored only in your result — automated detection reduces what reaches a provider, though it cannot guarantee elimination. Each deliberation leaves a timestamped record you can retrieve by ID. Retention periods and Zero Data Retention mode are set out in our Privacy Policy.

Does it replace associates, or our legal research tools?

No. It structures the opposition your position needs before it is exposed, which is a different job from finding the authority or drafting the document. Pilot5.ai does not provide legal advice. The research, the drafting and above all the judgment remain yours.

What does it cost to run on a live matter?

A deliberation is priced in credits and the estimate is shown before it runs, reserved at launch, settled on completion, with any excess refunded — so it fits inside a matter budget without a surprise on the bill. Current rates are on the pricing page.