Binding instruments
Primary law, regulations, directives, decisions and other binding instruments have a different evidentiary function from interpretive guidance or commentary.
Regulatory research fails when authority, applicability and effective state are flattened into one summary. A real document can still be non-binding. An official source can still be inapplicable to the jurisdiction or date that matters. Commentary can be useful without being authority.
Epistamate is designed to preserve those distinctions at the claim and evidence level, together with the audit trail of how the view was formed.
Primary law, regulations, directives, decisions and other binding instruments have a different evidentiary function from interpretive guidance or commentary.
Regulator guidance, Q&A, supervisory statements and official communications may be highly relevant while remaining non-binding. The distinction should stay visible.
Draft rules, consultations and legislative proposals are evidence about direction, not current obligations. Future-state and current-state claims should not silently merge.
Expert and industry analysis can orient research or identify contested interpretations, but it does not inherit the authority of the instrument it discusses.
Regulatory propositions can carry jurisdiction, authority class, binding status, affected actors, effective dates and applicability boundaries rather than losing those distinctions in synthesis.
Different instruments, regulators or jurisdictions can create genuinely conflicting or overlapping obligations. The product is designed to surface that state rather than forcing a false single answer.
Epistamate records research state and researcher decisions so a later reviewer can reconstruct what evidence and uncertainty were visible at the time. That can support governance and auditability; it is not a legal certification of compliance with any statute.
The long-term Knowledge layer is intended to preserve prior interpretations and evidence while re-evaluating them when an instrument changes, an effective date passes, or a new jurisdictional source appears.
Frontier models can identify missing jurisdictions, likely primary instruments, interpretive disputes and useful acquisition lanes faster than a rigid planner.
Models can help extract candidate obligations and relationships from long instruments, but candidate interpretation remains separate from evidence and legal authority.
Models can generate alternative readings, conflicting authorities and implementation edge cases. Those suggestions should drive evidence checks rather than become conclusions by agreement.
We are looking for bounded regulatory workflows where authority class, applicability and reviewer provenance are consequential.
Start a conversation →