AI & Output Disclosure
Effective date: May 1, 2026
Intrepid Prime is a legal intelligence company. Our product, KYRO, combines a proprietary federal procurement dataset with large language models and retrieval techniques to surface traceable answers to questions that would otherwise take attorneys hours or days to assemble.
This disclosure explains, in plain language, how KYRO works, what it is good at, what it is not good at, and how we think about the difference between KYRO Output and legal advice. We publish this document because customers and their clients deserve a clear account, and because we believe sophisticated buyers should understand the product they are paying for. This disclosure is not a substitute for the Master Services Agreement or Self-Serve Terms of Service (as applicable), the Acceptable Use Policy, or the Privacy Policy — in any conflict, those documents control.
1. What KYRO is
KYRO is a retrieval-grounded AI system that answers questions about U.S. federal contractors and federal procurement. At a high level:
- We ingest and normalize tens of millions of records from federal procurement sources, including the Federal Procurement Data System (FPDS), USASpending, and the System for Award Management (SAM).
- We structure this data across dimension tables covering contracts, obligations, agencies, contracting vehicles, set-asides, and related attributes.
- When you ask KYRO a question, the system retrieves the relevant underlying records and uses large language models to produce a structured, source-cited answer (“Output”).
- Output is designed to be traceable: every material factual claim aims to tie back to identifiable primary records. If the underlying records do not support a claim, KYRO is designed to say so rather than invent an answer.
Legal Intel Briefs are longer-form, templated Output products built on the same data and methodology.
2. How KYRO is different from a general-purpose chatbot
General-purpose chatbots answer questions using their training data plus, in some cases, live web search. They typically do not have access to the structured federal procurement record set, they cannot always cite their sources, and they are prone to confident-sounding errors on technical government-contracts questions — for example, confusing potential contract value with obligated revenue, or treating IDIQ ceilings as if they were committed spend.
KYRO is built to avoid those errors by grounding every answer in structured primary-source records, applying domain rules about how government contracts actually work, and exposing citations so a reviewer can check the math.
3. Known limitations
We would rather disclose our blind spots than have a customer discover them. The most important ones:
3.1 Other Transaction Agreements (OTAs)
A meaningful share of Department of Defense innovation spending now flows through Other Transaction Agreements and similar vehicles. OTAs are systematically underreported in FPDS and USASpending, and the underreporting is uneven across agencies and consortia. KYRO reflects the best available public data, but any analysis of OTA-heavy contractors will understate their total federal revenue to some degree. Where this is material to a question, KYRO flags it. If you are making an investment, M&A, or diligence decision that turns on OTA exposure, you should treat KYRO's numbers as a floor and supplement with direct Customer diligence.
3.2 Classified and unreported work
Classified contracts, certain intelligence-community work, and contracts below public-reporting thresholds do not appear in the source datasets we use. KYRO cannot surface what is not there.
3.3 Timing and latency
FPDS and USASpending data can lag actual contract actions by weeks or months, and corrections and deobligations are applied after the fact. Snapshot data from any given date is accurate only as of the most recent refresh.
3.4 Entity resolution
Contractors reorganize, rename, merge, and spin off. Parent-subsidiary relationships are not always clean in the public record. KYRO applies entity-resolution rules but will not always perfectly match every legal parent, subsidiary, or joint venture. When a question depends on precise corporate-family resolution, confirm with the underlying records.
3.5 Output variability
Large language models produce probabilistic Output. The same question asked twice may produce slightly different phrasing, different organization, or in rare cases different conclusions. Material factual claims should be stable across runs; stylistic presentation will vary.
4. Output is not legal advice
This is the most important point in this document. Intrepid Prime's founders are lawyers and we care about this deeply.
- KYRO Output is informational. It does not constitute legal advice, it does not create an attorney-client relationship, and it does not substitute for the judgment of qualified counsel applied to the facts of your matter.
- A law firm using KYRO is using an intelligence tool, in the same way a firm uses Westlaw, Lexis, or Bloomberg. The legal advice the firm provides remains the firm's advice.
- A corporate customer using KYRO for competitive intelligence, diligence, or compliance workflow is using an analytics tool. KYRO cannot take the place of engaging counsel for questions that actually require legal analysis (for example, a CFIUS filing decision, a disclosure judgment, or an export-control question).
- We do not provide legal services through the KYRO product. Where Intrepid Prime's affiliated attorneys provide legal services, they do so pursuant to a separate engagement letter, on legal services terms, and subject to the attorney-client privilege and the relevant rules of professional conduct.
5. How we handle Customer data used in KYRO
When you use KYRO, you may submit company names, UEIs, contract numbers, internal notes, and similar Customer Content as part of your queries.
- We use that content to answer your query and, in de-identified and aggregate form, to improve the product.
- We do not share your queries or Output with other customers.
- We do not sell Customer Content.
- We use third-party model providers to generate portions of Output. Queries and Output may be processed by those providers subject to their terms. We use enterprise-tier agreements with our model providers where available, which typically exclude our data from being used to train general-purpose foundation models, but we do not warrant that every model provider will maintain any particular practice in the future.
The full terms are in the Master Services Agreement. This disclosure is a plain-language summary, not a replacement.
6. Human review
We believe the right place to use KYRO is wherever the alternative is a blank page or a 40-hour billable block of contract-by-contract research. KYRO gets you to 80% faster than any other method we know of. We ask that you treat the last 20% — the judgment, the conclusions, the client-facing advice — as your work, informed by KYRO, not delegated to it.
In practical terms, this means: spot-check citations, understand what a number represents before you quote it, and do not paste KYRO Output into a regulatory filing without a professional having read it end-to-end first.
7. How to report issues
If you find an error in KYRO Output, a citation that doesn't resolve, or a pattern of mistakes in a particular query type, please tell us:
We read every report and prioritize fixes based on impact. Customer reports are the single most useful input into our quality improvement process, and this is a deliberate and central part of how the product gets better.
8. Changes to this disclosure
We will update this disclosure as the product and the underlying data landscape change. Updates take effect when posted to intrepidprime.com with a new effective date.