Pilot5 Legal positions itself as "deliberative AI designed for lawyers." It is not satisfied with giving fluent answers; instead, it builds answers around "work that can be challenged": claims come with authoritative sources alongside them, analytical inferences are labeled separately, opposing views remain in the record, and the reasoning process is reviewable.
Key capabilities Multi-model adversarial reasoning: five independent perspectives first analyze the question, then challenge one another, respond to objections, and revise under pressure. The final position follows the argumentative process, not the average of five first drafts. The strongest objection remains visible, and losing arguments are not erased. Sources separated from inferences: primary law is taken from publisher texts; analytical inferences are marked separately to avoid disguising confidence as authority. Every claim preserves a path back to the text. Reviewable record: every source, finding, challenge, and limitation is preserved together with the conclusion. The lawyer always remains the decision-maker and reviewer of the record. Five legal workflows
- Missing-clause checks in agreements: check separation agreements against selected statutory requirements and show the clauses behind each finding. Output an annotated Word agreement.
- Clause-by-clause contract interpretation: generate a clause-by-clause map of unfamiliar contracts, quote the original wording, flag skipped material, and help quickly locate the role of each numbered clause.
- Settlement pressure testing: apply pressure to both sides' positions, expose settlement ranges that impose costs on each party, and preserve five independent positions, statutory claim verification, and dissent.
- Verbatim statutory retrieval: paste citations and get publisher text without passing through a generative model. Supports up to 25 citations, provides official sources and retrieval dates, and leaves unresolved items unresolved.
- Firm position reuse: add approved firm positions to the statutory baseline used for future reviews, requiring human approval, preserving the baseline, and reusing approved positions in subsequent reviews.
How to use Choose one of the five workflows to start a task. The system shows a credit estimate before a paid run and refunds unused reserved credits after completion. There are no daily caps and no feature-gated plans. Monthly credits roll over across months for 12 months, and all plans include all five workflows. It can integrate with work environments such as Mistral · Le Chat, Microsoft Copilot Studio, Indian Kanoon, OpenCorporates, and Docket Alarm.
Use cases Suitable for legal work that needs to withstand challenge, such as contract review, settlement assessment, statutory verification, and standardization of firm positions. An interactive product tour can show every round of a completed Chapter 7 preference dispute from independent analysis to final synthesis.
Limitations Jurisdiction: each review is limited to U.S. law and does not consult other jurisdictions,Nor does it comment on other jurisdictions; this disclosure is passed along with each record of deliberation. Data and trust: Confidentiality obligations apply to client matters. Data is never used as a training asset, with no cross-account use and no model training. AES-256 encryption at rest, covering tables, WAL, and backups, and encryption cannot be disabled; the right to erasure with hard deletion is supported; retention periods are published by data category; frontier models run on zero-retention endpoints, and models without zero retention are excluded; pseudonymization before inference; TLS 1.3, AES-256-GCM, forward secrecy. Zero data retention mode, subprocessor disclosure, in-region deployment, BYOK and sovereign hosting, and a DPA under NDA (Article 28) may be requested. Fees and pricing: The official text does not provide specific pricing; it only states that capacity is selected by credit amount, and all plans include all workflows.