Many organizations document decisions and exceptions with procedures, approval workflows, ticketing systems and audit logs. That is exactly how it should be. But having an internal procedure is not the same as preserving the declared context of a decision in an independently verifiable record. The difference is where the record lives.
HSI's added value is not inventing yet another internal procedure. It is enabling the external deposit of a structured declaration, with a receipt time recorded by EVIDE, independently verifiable cryptographic integrity and, upon request, the completion of the evidentiary documentation through the FEDIS (Forensic Evidence Declaration & Integrity Statement): a technical document prepared for digital signature that identifies the deposited records and their cryptographic fingerprints, supporting their independent verification.
For smaller businesses, professional practices and independent professionals, who often lack structured internal documentation systems, this distinction can matter even more.
HSI does not replace governance, compliance procedures or audit systems: it adds a complementary evidentiary layer. EVIDE preserves what was declared; it does not determine whether the decision was correct, lawful or justified, nor does it verify the truth of the declaration.
With the EVIDE Mobile app you can deposit the declaration at the moment the decision closes: during a meeting, an inspection or a field activity, without waiting to get back to your desk.
Once authenticated, EVIDE Mobile can retain your authenticated session on the device, reducing repeated login steps when you return to the app.
Search, filter, open and review your previous EVIDE intakes directly from your phone, with their evidentiary status visible at a glance.
Human Structured Intake lets you record a structured, attributable human declaration, decision or intervention directly from the mobile environment, particularly useful during meetings, inspections or other field activities.
EVIDE Mobile launches in English, Italian, German, French and Spanish. Bringing it to five languages required enormous care: evidentiary terminology cannot simply be translated word for word - identity, status, declarations, interventions and case relationships must remain clear and consistent across the entire mobile experience.
Case Navigator lets you follow the registered relationships between linked intakes and certifications, node by node, from your phone, one of the more distinctive things EVIDE Mobile does.
Share an HSI record with a person named individually and review the summary of responses. Anyone who receives an invitation opens it by pasting the link and deposits their position directly from the app.
Swipe through the screens to see how a structured declaration is completed in the EVIDE Mobile app, how the record is shared for a response and how the invited person responds. Each section of the form matches the fields described in this guide.
The app is available in 5 languages: English, Italian, French, German and Spanish.


















How an HSI declaration becomes a preserved record, and which elements it contains. Each field is explained in detail below.
The depositor of an HSI record can invite other people by name to declare their position on that record, on the web and in the EVIDE Mobile app. Each response is a distinct record with its own EAR-LINKED; on request, a forensic expert documents the record and the responses in a signed FEDIS-LINKED.
On request, a forensic expert examines the HSI record and issues the FEDIS (Forensic Evidence Declaration & Integrity Statement), a document distinct from the deposit.
The FEDIS is signed by the certifier with a qualified electronic signature.
It receives a qualified electronic time stamp (Article 41 of the eIDAS Regulation) issued by a third-party qualified trust service provider: the document carries a date that can be relied upon against third parties.
It reports the SHA-256 fingerprints of the record and of its EAR, which make it possible to verify that the record existed in that form no later than the date of the time stamp.
For a record with linked responses, the FEDIS-LINKED also documents the responses, with their fingerprints and the integrity checks performed at generation.
💡 Requesting the FEDIS right after the deposit brings the certain date closer to the moment of the decision.
The time stamp applies to the FEDIS document: it does not date the original decision and does not make the receipt time of the deposit qualified. The FEDIS attests the existence and integrity of the record, not the correctness of the decision or the truth of what was declared.
HSI is for people who, after a site inspection, an audit or a meeting, take or put in writing a decision that may be questioned months later. Here are the roles it is designed for, and what it preserves for each of them.
Deposit the decision to notify or not notify a breach, with the declared grounds and uncertainties. The register stays in the organization: HSI refers to it by its SHA-256.
After a site inspection, deposit the assessment and the measures decided, together with what could not be verified. The employer and worker safety representatives can declare their own position with Linked Responses.
Deposit from your phone, on site, the instruction given to contractors. The contractor and the works director can declare whether they agree or dispute it.
Deposit the finding and the recommendation, with the scope of what was verified and what was not. Management can acknowledge or dispute it.
Preserve case-level compliance decisions outside internal systems, with the declared role and the receipt time recorded by EVIDE.
Deposit the decisions taken during an incident or an assessment, such as isolating, keeping a service running or escalating, with references to tickets and alerts that stay in your systems.
Deposit the nonconformities found and the closure of the audit. The audited organization can acknowledge or dispute them, and each position remains distinct.
Deposit the findings of the inspection and the measures decided, with the declared limitations, such as inaccessible areas or missing documentation.
Deposit the assumptions of the energy audit and the data that were not measured directly, so the basis of the conclusions remains clear.
Deposit the assessments made during site visits and the decisions taken after an event in the transport of dangerous goods, with the receipt time.
Deposit the corrective actions decided after an inspection, including the decision to withdraw or not withdraw a product, with the uncertainties of that moment.
Deposit the decisions on waste, discharges and emissions taken after an inspection, with references to documents that stay in the organization.
Deposit the assessment made after inspecting the loss. The insured party and the broker can declare their own position on the same record.
Deposit the decision taken urgently or at a meeting. Invited owners can declare whether they agree, dispute or acknowledge it.
Deposit the handover of a plant, a system or a piece of work. The recipient declares their own position, and reservations remain preserved.
Can advise clients to deposit relevant decisions when they are taken, so that any later dispute rests on what was declared at the time rather than on a later reconstruction.
Deposit decisions on measures and reports, with the declared role. The people involved can deposit their own account as a distinct record.
Deposit the human decision to approve, stop or correct an AI system, distinct from the system’s own logs.
In every case, HSI does not replace minutes, signatures, registers or legal obligations. It deposits the declared decision externally, with the receipt time recorded by EVIDE, and does not establish whether the decision was correct.
Most organizational decisions leave a trace somewhere, an email, a ticket, a comment in a workflow tool, a line in a spreadsheet. That trace is rarely structured, rarely time-anchored in a tamper-evident way, and rarely designed to survive outside the system that created it. When a decision is later questioned, in an audit, a dispute, a regulatory inquiry, or an internal investigation, reconstructing exactly what was decided, by whom, in what declared role, and under what conditions can be slow, incomplete, or impossible.
Human Structured Intake gives a person a structured way to declare a decision at the moment it closes: the type of decision, who declared it and in what role, when it was closed, the applicable classification and threshold context, whether a readiness gate was involved, and any signals that remained unresolved. EVIDE then canonicalizes that declaration, computes a deterministic hash over it, and preserves it as an independently retrievable record.
This distinction runs through every field documented on this page. Where a value uses a word like «verified,» that word describes what a person or an upstream mechanism declared, not an assessment performed by EVIDE itself, unless stated otherwise for that specific field.
People evaluating Human Structured Intake usually compare it with a tool they already use: a register, a logging system, an approval workflow, a set of minutes. Below are the objections that come up most often, role by role, each with its answer and the boundary of what EVIDE claims.
None of these tools is replaced. HSI adds a different layer: the external deposit of the declaration about a single decision, with the receipt time recorded by EVIDE and integrity verifiable through SHA-256.
Seven cases, seven videos. Each enterprise use case on this page is also illustrated in a short video.
Watch the seven cases on LinkedIn ↗The register is the right foundation and HSI does not replace it, just as it does not replace the documentation duties set by the GDPR. The register, however, is kept and updated inside the organization. HSI deposits externally the declaration about a single decision, for example, to notify or not to notify, with the receipt time recorded by EVIDE, distinct from the declared closure time (Decision Closure). The deposit can refer to the register entry with its declared SHA-256, without the register ever leaving the organization (External Evidence References).
Logs record what systems did; HSI preserves the declared decision, keep a service running, isolate it, escalate, with the declared decision-maker and role, kept distinct from the verified DAPI identity of the person depositing the record, and the declared basis. Logs and tickets stay where they are: the HSI record refers to them through the ticket or alert ID (Trace Reference), without transferring technical data to EVIDE. What the declarant still considers uncertain is recorded as a declared uncertainty; unresolved signals belong only to a separate verification mechanism.
Minutes document the meetings of the oversight body and are often drafted or approved at a later time. HSI concerns the single operational decision at the moment it closes, for example, an exception authorized by a business function, with the declared role, the reference procedure or threshold (Threshold Reference) and the two distinct times of closure and receipt. An oversight body can ask for relevant decisions to be deposited, and examine them later without depending on the systems of the function that made them.
The workflow records that the approval happened. HSI preserves the declared context of that approval: the reference policy and how clearly its ownership is attributable (Threshold Attribution), whether a separate mechanism assessed the conditions (Boundary Readiness) and what remained uncertain. And it preserves it externally, with integrity verifiable through SHA-256.
That is correct: HSI preserves a declaration, it does not make it true. What it adds is different: it makes it possible to show that the declaration existed in that form at the receipt time recorded by EVIDE, and that it was not reconstructed after that moment. Upon request, the FEDIS identifies the deposited records and their cryptographic fingerprints, in support of their independent verification. The weight given to the declaration remains a judgment for whoever examines it.
No. HSI is a tool for the external evidentiary preservation of declarations about specific decision moments. EVIDE does not follow the process in real time and does not verify that the declared activities were actually carried out. For each deposit it records the moment of receipt, distinct from the declared decision time, and preserves its cryptographic integrity. Later decisions are documented with new linked deposits, without modifying earlier declarations (Human Judgment Evolution): the result is a sequence of preserved declarations that can be examined later.
The system records the outcome. HSI makes it possible to keep what the automated system recommended separate from the declared human decision, with the declared decision-maker and role, kept distinct from the verified DAPI identity of the person depositing the record, the declared closure time and the elements declared as uncertain: a distinction that is hard to reconstruct afterwards from logs alone. The record can refer to the case file without containing unnecessary personal data.
The system tracks that the override happened and under which account. HSI preserves, outside that system, the declared basis of the override, the declared status of the risk threshold and what remained uncertain. It is not meant for every transaction: it is meant for the overrides you will want to be able to reconstruct.
The checklist documents the checks performed. HSI preserves the declared restart authorization, with the declared decision-maker and role, kept distinct from the verified DAPI identity of the person depositing the record, the declared diagnostic check and its scope (Gate Identifier, Gate Scope), and what remained uncertain. The checklist can be referenced with its declared SHA-256, without being uploaded.
The platform records the approval within itself. HSI preserves the declared authorization externally: who is declared as authorizing and in what role, kept distinct from the verified DAPI identity of the person depositing the record, the declared conditions and what remained uncertain. The record remains available even if the platform changes or its logs are rotated.
No one can rule that out, which is exactly why HSI distinguishes two times. The closure time is declared by the person recording; the receipt time is recorded by EVIDE and cannot be chosen by the depositor. A deposit made after the event shows up as such, and a long gap between declared closure and receipt is immediately visible to whoever examines the record (Decision Closure).
The decision’s closure time is declared separately: the deposit can happen later, once things have stabilized, without confusing the two moments. With the EVIDE Mobile app, it can also be deposited in the field, during a meeting or a site visit.
With an external reference, the document is not sent: you declare a reference and, if you wish, its SHA-256 (External Evidence References). EVIDE receives the content of the declaration, and the depositor chooses what to write, using references instead of data where needed.
A Human Structured Intake turns a human decision into a structured evidentiary object. Here is exactly what that means, in three steps.
What You Declare Declared by the user
Title, Business area involved, Description, Internal contact, Case reference, Evidence reference, Additional notes.
Decision type, Decision closure date/time (entered in local time and converted to UTC before submission).
Role of the person who made the decision, Name of the person who made the decision, Reference on the decision-maker’s authority (optional).
Classification stability (optional), Threshold reached (optional), Threshold reference, Threshold attribution clarity (optional).
Trace reference (optional), Decision readiness, Identifier and scope of the separate verification mechanism, Signals not resolved by the verification mechanism, Declared uncertainties (optional).
Document / external artifact - Artifact type, Reference, Declared origin, Relationship to the decision, Description, Retention status, SHA-256 of the document, Who computed the hash, Hash scope.
What EVIDE Preserves
| Preserved evidentiary object Preserved by EVIDE |
| Server-generated Intake Timestamp Server-recorded |
| SHA-256 integrity hash Server-computed |
| DAPI identity reference Preserved reference |
DAPI identity reference: reference to the independently verified identity associated with the intake.
Preserved as received. Each deposited state remains a separate evidentiary object, later states do not rewrite previously deposited states.
What EVIDE Derives
A server-computed signal describing how consistent the declared classification is with what was observable at the evidentiary boundary.
A server-computed signal describing the observational quality of the conditions under which this single closure was stabilized, not a measure of decision volume or throughput.
A server-computed signal describing whether declared accountability remains structurally supported, based on this closure’s authority, threshold attribution, and continuity state.
Derived signals are computed separately from the declared evidentiary object. They do not rewrite the original intake.
The seven scenarios below are illustrative examples of how different business functions might use Human Structured Intake. Each shows the kind of decision involved, information that could reasonably be declared, the regulatory context in which that decision may become relevant and, just as importantly, the boundary of what EVIDE’s preservation of that record does and does not establish.
Regulatory references. Each case indicates the rules that may make the documented decision relevant: the AI Act, Italian law and, where pertinent, other frameworks (NIS2, GDPR, machinery safety). The reference describes the context; it does not mean that an HSI record satisfies those obligations, verifies compliance with them or attests conformity. Many AI Act obligations for high-risk systems apply from 2 December 2027 or from 2 August 2028: dates, conditions and exceptions are set out on the dedicated page.
EVIDE & AI Act - regulatory framework ↗Seven cases, seven videos. Each enterprise use case on this page is also illustrated in a short video.
Watch the seven cases on LinkedIn ↗Event / ContextA Compliance Officer is asked to temporarily authorize an exception to a procedure that would normally be blocked, under time pressure and with a documented business justification.
Human DecisionThe officer approves a temporary, scoped exception, subject to specific conditions and a defined review point.
Information That Could Be Declared
Reconstructive ValueA later reviewer, auditor or regulator can reconstruct that the exception was declared, by whom, in what role, under what declared threshold and readiness conditions, without relying on the officer’s memory or on a system log that may since have been rotated.
Event / ContextA Security Operations Center detects anomalous behavior. A human responder reviews the available telemetry and must decide whether to allow continued operation, block, isolate or escalate.
Human DecisionThe responder authorizes continued operation while the investigation proceeds, rather than an immediate isolation.
Information That Could Be Declared
provisional, since the investigation is not yet closedcandidate if only the responder carried out the assessment; verified_partial only if a separate verification mechanism (for example an independent tool or team) assessed the conditions with incomplete visibility, stating its identifier and scopeverified_partial or unverifiable: what the separate verification mechanism could not resolve, e.g. «Origin of two API requests not yet attributed»declared_uncertainties), with any readiness value, including candidate: what the responder personally still considers uncertain, e.g. «Possible EDR sensor false positive»Reconstructive ValueIf the incident later escalates, the record preserves what the responder declared as known, verified or unresolved, with the closure time the responder declared and, separately, the time at which EVIDE received the record. This helps distinguish hindsight from what had been declared at the time; it does not prove that the declared information was complete or accurate, nor that the decision was made exactly at the declared time.
verified_partial with the unresolved signals written out in full is more honest than a premature verified. If no separate mechanism was involved, the correct value is candidate: a personal assessment should not be presented as independent verification. With candidate, whatever remains uncertain is recorded as a declared uncertainty, not as an unresolved signal: signals belong to a separate mechanism and, with candidate, the request would be rejected. In both cases this is a declaration: EVIDE does not verify that the gate exists or is independent.Event / ContextAn AI-assisted system produces a recommendation about a candidate or employee. A human reviewer examines that recommendation and makes the final decision.
Human DecisionThe reviewer accepts, modifies or overrides the automated recommendation.
Information That Could Be Declared
Reconstructive ValueThe record helps preserve a structural separation between the recommendation attributed to the automated system and the declared human decision, a distinction that is often difficult to reconstruct after the fact from system logs alone.
Event / ContextA fraud prevention system blocks a transaction as high risk. A fraud analyst reviews the case and must decide whether to release it, keep it blocked or escalate it.
Human DecisionThe analyst releases the transaction after review, judging the risk signal to be a false positive.
Information That Could Be Declared
Reconstructive ValueThe record preserves the declared context of the human override, who authorized the release, when and on what declared basis, regardless of whether the transaction later turns out to have been legitimate or not.
Event / ContextA production line stops automatically after an anomaly is detected. A technical manager reviews the relevant conditions and must decide whether to authorize a restart.
Human DecisionThe manager authorizes the restart, having reviewed the available diagnostic information.
Information That Could Be Declared
verified_partial - e.g. «Sensor 4 reading not yet cross-validated»; if only the manager carries out the assessment (candidate), the same uncertainty is declared as a declared uncertaintyReconstructive ValueIf a related failure recurs later, the record preserves which diagnostic check was declared as performed, with what declared scope, in the declaration authorizing the restart.
verified_partial, with the identifier and scope of the check and the unresolved signals, rather than a full verified. If the assessment is the manager’s alone, the correct value is candidate.Event / ContextAn AI agent requests to perform a sensitive action. Before execution, a human authority must approve, reject, condition or escalate the request.
Human DecisionThe human authority approves the requested action, subject to declared conditions.
Information That Could Be Declared
Reconstructive ValueHuman Structured Intake records the human decision to authorize. It is deliberately not the same function as declaring the operational perimeter within which the agent was authorized (that is the separate declarations extension documented in the JSON Schema Reference), and it is not an automatic execution-authorization mechanism.
Event / ContextDuring a cybersecurity incident, a responder must choose between competing courses of action under time pressure, for example isolating a server versus keeping a service operational despite an indicator of compromise.
Human DecisionThe responder declares a chosen course of action, together with the information the responder declares available at that moment.
Information That Could Be Declared
provisional during an active incidentverified_partial or unverifiable)Reconstructive ValueIncident timelines are frequently reconstructed after the fact under scrutiny. A structured, time-anchored declaration reduces reliance on reconstructing intent from memory or from logs that were not designed for evidentiary use.
candidate and whatever the person still considers uncertain is declared as a declared uncertainty. Unresolved signals belong only to a separate verification mechanism (verified_partial or unverifiable). Neither is verified by EVIDE.These cases concern areas in which the AI Act, or Italian law, gives specific weight to the human decision taken with the support of an AI system. In each of them, the question that comes afterwards is the same: what did the system produce, what did the person decide, and what had been declared as known, verified or unresolved at that moment.
An HSI record documents the decision as declared. It does not prove that human oversight was effective, it does not classify the system as high-risk and it does not attest conformity with the AI Act. Whether a system actually falls within one of these areas depends on its intended purpose and on Art. 6(3) of the AI Act, and must be assessed case by case.
Event / ContextAn AI system computes a credit score, or a risk assessment and pricing for life or health insurance. An operator examines the outcome before the decision.
Human DecisionThe operator confirms, modifies or rejects the outcome proposed by the system.
Information That Could Be Declared
candidate if no separate mechanism has assessed the conditionsdeclared_uncertainties), e.g. «Declared income not yet documented»Reconstructive ValueThe record keeps the system’s outcome and the operator’s decision distinct, with the context declared at the moment of closure: exactly what is needed when one has to explain what role the system played in the decision.
Event / ContextA remote biometric identification system flags a match. Before any action, the match must be separately verified and confirmed by at least two persons.
Human DecisionEach verifier confirms, or does not confirm, the match.
Information That Could Be Declared
declared_uncertainties), e.g. «Image quality insufficient for a full comparison»Reconstructive ValueTwo separate records, each with its own Intake Timestamp and the verified DAPI identity of the session recording it, show that two confirmations were declared, by which session each was recorded and in what order EVIDE received them. Two declarations, two DAPI identities and two timestamps do not in themselves prove that two genuinely separate and independent verifications took place, nor that they took place before the action: the DAPI identity indicates who entered the record, the closure time is declared and the Intake Timestamp indicates only when EVIDE received each record.
Event / ContextAn AI system grades a test, proposes an admission outcome or flags anomalous behavior during an exam. A teacher or a board examines the result.
Human DecisionThe teacher or the board confirms, modifies or annuls the proposed outcome.
Information That Could Be Declared
candidate if no separate mechanism has assessed the conditionsdeclared_uncertainties), e.g. «Flag of anomalous behavior not confirmed»Reconstructive ValueIf the outcome is challenged, the record shows what the system proposed according to the declaration, what the person declared to have decided and which elements were declared open at the moment of the declared closure.
Event / ContextAn AI system supports the assessment of eligibility for a benefit or an essential public service. The responsible official examines the file.
Human DecisionThe official adopts the administrative decision, either confirming or departing from the system’s indication.
Information That Could Be Declared
candidate if no separate mechanism has assessed the conditionsdeclared_uncertainties), e.g. «Income documentation pending verification»Reconstructive ValueThe record documents the official’s decision as an act distinct from the system’s output, with the context declared at the moment of closure.
Event / ContextAn AI system classifies emergency calls and proposes a response priority or the dispatch of a unit. The control-room operator assesses the proposal.
Human DecisionThe operator confirms or modifies the priority and the dispatch.
Information That Could Be Declared
candidate if no separate mechanism has assessed the conditionsdeclared_uncertainties), e.g. «Caller’s location not confirmed»Reconstructive ValueDistinguishing the declared moment of the decision from the moment of recording makes it possible to reconstruct the sequence even when the record was entered at the end of a shift.
Event / ContextAI software that is a medical device proposes a diagnostic hypothesis or a reporting priority. The physician examines the proposal.
Human DecisionThe physician confirms, modifies or discards the system’s indication.
Information That Could Be Declared
candidate if no separate mechanism has assessed the conditionsdeclared_uncertainties), e.g. «Previous comparison exam not available»Reconstructive ValueThe record shows that the clinical decision was declared by the physician as an act distinct from the system’s indication, with what the physician declared as known and as still open at the moment of the declared closure.
A decision rarely concerns one person only. Whoever took it needs to know, and often to be able to show later, what the others thought of it: the colleague who was present, the client, the supplier, the person called into question, the advisor. Today those positions stay scattered across emails, messages and conversations, written at different times and about different versions of the facts.
With Linked responses, the depositor of an HSI record invites a person by name to declare their position on that record. The person reads a representation of the record frozen at the time of the invitation, chooses one of four positions, can give reasons, and deposits the response with their own EVIDE account. EVIDE records the receipt time and generates an EAR-LINKED document for each response. At the request of the depositor of the record, a forensic expert can document the record and the responses in a signed FEDIS-LINKED.
The value does not lie in proving anyone right. It lies in the fact that each position remains a distinct record, attributed to the account that deposited it, with its own time and fingerprint, and that no deposit modifies or deletes anyone else's.
| Position | What the respondent declares |
|---|---|
| I AGREE (CONCORDO) | Agreement with what the representation reports. Reasons and reservation are optional. |
| I ACKNOWLEDGE WITHOUT EXPRESSING AGREEMENT (PRENDO ATTO SENZA ESPRIMERE ACCORDO) | Having taken note of what is reported, without stating agreement. It is the typical position of someone receiving a handover or a piece of information. |
| I DISPUTE (CONTESTO) | Disagreement. Reasons are required; a reservation cannot be combined with this position. |
| I AM NOT IN A POSITION TO ASSESS (NON SONO IN CONDIZIONE DI VALUTARE) | Not having the elements to express a view. It is neither agreement nor dispute. |
The four positions carry equal weight and are fixed Italian labels: in the other languages they appear as a translation followed by the original in brackets, because the declared and preserved value is the original.
B always sees their own responses, in the «My responses» section. A sees them as long as A has access to the record. For an organization record, A can choose at the time of the invitation that the response is also visible to the current members of the organization; former members do not see it. B reads this indication before depositing, and the visibility stays the same for all later responses of the same person on the same representation.
A can revoke an invitation; for organization records, revocation may follow a procedure with a request and a decision by a different owner or admin. Revocations, suspensions and expiry prevent new responses, but do not modify or delete those already deposited.
The FEDIS-LINKED is the FEDIS of an HSI record that documents, in addition to the record, the includable linked responses. It is requested by the depositor of the record, from the record's FEDIS button on the web, with the same procedure and the same financial terms as any FEDIS: the certifier's quote, express acceptance, any FEDIS packages. Nobody else can request it on that record, not even the other members of the organization.
On request, the platform forms from its own archive a snapshot of the relevant records, identified by its SHA-256 fingerprint. For each included response, the document reports the respondent's name and DAPI number, any declared role, the position, the verbatim reasons and reservation, the deposit date, the visibility indicated, the identifiers and fingerprints of the response and of its EAR-LINKED, and descriptive counts of invitations and responses. For non-includable responses and invitations without a deposit it reports only counts, without names or content. At generation, the platform recomputes the fingerprints of the records still available: if one does not match, the document is not generated.
A forensic expert examines the document and signs it. Later requests form a series: each FEDIS-LINKED documents the records present at the date of its own snapshot, cites the previous emissions and does not modify, replace or withdraw any document already issued. Only responses deposited after the respondent was shown the FEDIS-LINKED notice are includable: the respondent knows in advance that their position may end up in a document handed to third parties, even if on the platform the response is visible only to A.
The complete rules for Linked responses and FEDIS-LINKED are in Sections 7.7 and 7.8 of the Service Conditions.
Read the Service Conditions ↗In each of these cases the HSI record remains the declaration of its depositor; the responses add the positions declared by other people, each with its own provenance. All recipients need an EVIDE account with a DAPI number.
The recordA, the department head, deposits an HSI record: decision type, for example «Reconstruction of the line 3 stoppage», declared role, closure time, a reference to the incident ticket and, as a declared uncertainty, what A could not establish. In A's reconstruction the cause is an operation carried out by C.
The responsesA invites B, a colleague present at the time of the stoppage, who declares I AGREE (CONCORDO). A can also invite C: in that case C can declare I DISPUTE (CONTESTO) with their own reasons, which remain preserved alongside the other positions.
C's autonomous recordWhether invited or not, C can deposit their own HSI record with their version of the facts, cite the EVIDE ID of A's record as a declared reference, if C knows it, and in turn invite other people. C is not forced to defend themselves inside A's account, and A's record does not become the official version just because it was registered first.
Reconstructive ValueThe organization does not have to decide straight away who is right. It can reconstruct what A declared and when, what B declared, what C declared independently, which evidence was cited and which was declared missing. B's agreement does not in itself prove C's responsibility. At A's request, the FEDIS-LINKED can document A's record with the includable responses; C can request a FEDIS on their own record.
The recordAfter a leak from the roof, the administrator orders urgent works. The administrator deposits an HSI record with the decision type, for example «Urgent roof repair», the declared role of administrator, the closure time, the quote as an external reference with its declared SHA-256 and, as declared uncertainties, what was not yet known, such as the extent of the damage.
The responsesThe administrator individually invites the members of the building council, or the owners who have an EVIDE account. Each declares their position on the same frozen representation: I AGREE (CONCORDO), I DISPUTE (CONTESTO) with reasons, I ACKNOWLEDGE WITHOUT EXPRESSING AGREEMENT (PRENDO ATTO SENZA ESPRIMERE ACCORDO) or I AM NOT IN A POSITION TO ASSESS (NON SONO IN CONDIZIONE DI VALUTARE).
Reconstructive ValueWhen reporting to the meeting, the administrator has their own declared decision and the individual positions collected in the days of the works, each with its own receipt time.
The recordA loss adjuster, an installer or a party-appointed technician carries out an inspection in the presence of the owner, the contractor and an insurance representative. The technician deposits an HSI record with the declared outcome, the photographs as external references with their SHA-256, without handing them to EVIDE, and, as declared uncertainties, what could not be inspected.
The responsesThe technician invites each of the parties present. The owner declares I AGREE (CONCORDO), the contractor I DISPUTE (CONTESTO) with its reasons, the insurance representative I AM NOT IN A POSITION TO ASSESS (NON SONO IN CONDIZIONE DI VALUTARE) pending documentation.
Reconstructive ValueThe parties' positions are collected on the same representation of the outcome, shortly after the inspection, each with its own provenance. Any disagreement stays documented with its reasons, instead of surfacing months later as conflicting recollections.
The recordThe person responsible for the decision deposits an HSI record: notifying or not notifying the breach to the supervisory authority, with the declared role, the closure time, the reference to the breach register entry with its declared SHA-256, and the declared uncertainties.
The responsesThey invite the CISO, the DPO or the legal advisor. Each declares and explains their position; a dissenting view stays preserved as I DISPUTE (CONTESTO), with its reasons.
Reconstructive ValueOver time it is possible to reconstruct not only the declared decision, but also which functions were consulted and what position they had declared at that moment.
The recordAt the end of the shift, the outgoing supervisor deposits an HSI record with the declared state of the plant, the open anomalies as declared uncertainties and the reference to the shift log.
The responsesThe outgoing supervisor invites the incoming shift supervisor, who declares I ACKNOWLEDGE WITHOUT EXPRESSING AGREEMENT (PRENDO ATTO SENZA ESPRIMERE ACCORDO), or I DISPUTE (CONTESTO) with reasons if the situation differs from what was declared.
Reconstructive ValueIf it later has to be established what was communicated at the shift change, what the outgoing supervisor declared to hand over and the position declared by the incoming one remain distinct.
The recordThe supplier's service manager deposits an HSI record with the declared closure of the outage, the declared restoration time and the reference to the ticket.
The responsesThe service manager invites the client's contact person, who belongs to another organization and has no access to the supplier's systems. The contact reads only the FAST REVIEW and declares I AGREE (CONCORDO), or I DISPUTE (CONTESTO) stating, for example, that the service was not yet fully restored.
Reconstructive ValueBoth positions stay preserved outside the systems of both parties, without the client having to receive the complete record or the supplier's internal documents.
The recordA consultant, a lawyer or an accountant deposits an HSI record with the declared advice, the choice communicated by the client, the closure time and the risks stated as declared uncertainties.
The responsesThe professional invites the client, who declares I AGREE (CONCORDO) if the representation matches what was agreed, I ACKNOWLEDGE WITHOUT EXPRESSING AGREEMENT (PRENDO ATTO SENZA ESPRIMERE ACCORDO), or I DISPUTE (CONTESTO) with their own reasons.
Reconstructive ValueIf a dispute later arises about what had been advised, the professional's declaration and the client's declared position on the same representation remain preserved.
The value of Linked responses lies in what they preserve. Their boundaries are what make that value reliable for whoever examines the records later.
Two different timestamps appear in every Human Structured Intake record, and they answer two different questions:
| Timestamp | What it answers |
|---|---|
decision.closure_timestamp_utc | When was the decision itself closed? This is the moment at which the person declares the decision was actually made, entered by the person, in UTC. |
| Intake Timestamp | When did EVIDE receive and preserve this record? This is generated by the server at the moment the intake is processed, not entered by the person. |
These two moments are frequently different, and that difference is expected, not an error. A decision might be closed at 16:00 and only recorded in EVIDE at 18:03, because of a delayed workflow step, a batch submission process, or simply because the person recording it does so later in the day.
decision_closure_timestamp_utc is a value declared by the person completing the intake. It is not a cryptographic or «trusted» timestamp of when the decision was actually made in the real world - EVIDE has no independent way to confirm that the declared closure time is accurate. What EVIDE does preserve, independently, is the Intake Timestamp: the moment the record itself was received and canonicalized.Human oversight is frequently represented in governance tooling as if it occurred at a single moment: a recommendation is produced, a person reviews it, and a final decision is recorded.
In practice, human judgment may evolve. New information can arrive after an initial recommendation but before a later accountable decision, and the reviewer’s assessment, rationale, classification or decision posture may change as that information becomes available.
A single Human Structured Intake record preserves one declared state at the time of intake. Where multiple states need to be preserved over time, a sequence of linked Human Structured Intake records can preserve those successive declarations independently. Each state remains its own evidentiary object, with its own server-generated Intake Timestamp and its own hash computed over that record’s evidentiary payload, a later state does not rewrite, normalize or replace an earlier state.
| 1. Initial declared review state |
| 2. Additional information received |
| 3. Reviewer’s assessment changes |
| 4. Further review or escalation |
| 5. Later declared decision state |
Each step may be preserved as an independent Human Structured Intake record. The structural relationship between deposited states can be reconstructed by an authorized consumer without modifying any previously deposited state.
What this does not establish
The eight fields below provide operational and organizational context. They are shared with EVIDE’s other intake types and are not specific to the Human Structured evidentiary model, they are documented here briefly, for completeness, rather than in the same depth as the fields in the next section.
| Field | Type | Requirement | What it is |
|---|---|---|---|
title | string | Required | Short label for the record. Verified in the code: this value is copied directly into decision.summary in the canonical payload, it is not a separate, unrelated field. |
business_area_id | integer | Required | The business area the intake belongs to. Verified in the code: must reference an existing, active business area, otherwise the request is rejected. |
description | string | Optional | Free-text description stored alongside the record. Not included in the canonical evidentiary payload itself. |
context_reference | string | Optional | Free-text operational reference for internal use. |
notes | string | Optional | Free-text internal notes. |
topic | string | Optional | Free-text topic label for organizing intakes. |
case_reference | string | Optional | Free-text case or matter reference. This is the same field used by the «Evidence from the same case» search available elsewhere in EVIDE. |
evidence_reference | string | Optional | Free-text evidentiary reference for internal cross-linking. |
These fields provide operational and organizational context for locating and managing the intake within the platform. They are not part of the Human Structured evidentiary model documented in the rest of this page, and their presence or absence does not affect the evidentiary fields below.
The fifteen fields below are the fields specific to the Human Structured evidentiary model, verified individually against the EVIDE validator. The table gives a quick summary; a full explanatory card follows for every field.
| Field | Type | Requirement |
|---|---|---|
decision_type | string | Required |
decision_closure_timestamp_utc | datetime | Required |
authority_role | string | Required |
authority_id | string | Optional |
authority_verification_note | string | Optional |
classification_status | enum | Optional |
threshold_status | enum | Optional |
threshold_reference | string | Conditional |
threshold_attribution_status | enum | Optional |
trace_access | string | Optional |
boundary_readiness_status | enum | Required |
readiness_gate_identifier | string | Conditional |
readiness_gate_scope_reference | string | Conditional |
unresolved_signals[] | array of strings | Conditional |
declared_uncertainties[] | array of objects | Optional |
decision.type
A free-text label describing the category of the human decision or intervention being recorded, for example, what kind of decision this is, not what the decision concluded.
Without this field, a record would not indicate at all what kind of decision or intervention was recorded. It gives every other field in the record a category to refer to.
Always, this field is required for every Human Structured Intake.
Choose a label specific enough to distinguish this decision from other decisions of a different kind, but do not duplicate the full narrative that belongs in the title field (which becomes decision.summary). A good decision type names the category of action; the title/summary describes this specific instance.
Distinct from the common field title, which is copied directly into decision.summary - a short natural-language description of this specific decision. decision_type names the category; decision.summary narrates the instance.
A Compliance Officer approving a temporary policy exception might record the decision type as «Compliance exception approval.» A technical manager authorizing a production restart might use «Production line restart authorization.»
decision.closure_timestamp_utc
The declared moment at which the human decision itself was closed, not the moment the record was submitted to EVIDE. See the dedicated Decision Closure section above for the full distinction from the Intake Timestamp.
A decision and its evidentiary preservation rarely happen at the exact same instant. Recording the declared closure time, separately from the server-generated intake time, makes it possible to distinguish when the decision was declared closed from when EVIDE received the record. The closure time remains a declared value: it does not by itself prove when the decision took place in the real world.
Always, this field is required. Enter the moment the decision was closed, not the moment you are filling in the form.
In the web form, enter your device’s local time: the browser converts it to UTC and shows, before submission, the UTC time that will be recorded together with the time zone used. In the web form, a time that does not exist in your time zone (switch to daylight saving time) is rejected; a time that occurs twice (return to standard time, for example 02:30 on 26 October 2025 in Italy) requires you to choose the offset explicitly, with no automatic choice. The mobile app performs the same conversion to UTC; the explicit choice for repeated times is currently available only in the web form. Via the API, the value must be sent in ISO 8601 with Z (for example 2026-04-07T16:00:00Z); an explicit offset such as +02:00 is also accepted and converted exactly to UTC.
YYYY-MM-DDTHH:MM with optional seconds, followed by Z, by an offset ±HH:MM or by no zone, interpreted as UTC). The following are rejected with an explicit error, never silently corrected: relative expressions such as «yesterday», non-existent dates or times (30 February does not become 2 March), closures more than 5 minutes later than the server time and closures earlier than 1 January 2000. An empty value is rejected, rather than falling back to the current time.A production line stops at 14:32 local time. The technical manager reviews the conditions and authorizes a restart, closing that decision at 16:00 local time, but only records the intake in EVIDE at 18:03, after finishing other urgent tasks. The closure timestamp preserves 16:00 (converted to UTC); the Intake Timestamp separately preserves 18:03.
Z) produces a closure shifted by your own offset: a value that falls in the future relative to the server is rejected, one in the past is not. In the web form and in the app, the conversion is automatic and visible before submission.authority.role
The organizational role, in what capacity, the decision is declared to have been made, distinct from the identity of the person who made it.
A name alone does not indicate in what organizational capacity a decision was taken. The role helps a later reader reconstruct the declared organizational authority behind the decision, regardless of who specifically held that role at the time.
Always, this field is required for every Human Structured Intake.
Use the role title that is organizationally meaningful for this decision, e.g. «Compliance Officer,» «SOC Manager,» «HR Director,» «Fraud Analyst,» «Plant Manager.» Free text, not a fixed list.
Distinct from authority_id (who) and from the DAPI identity of the logged-in session (see Authority Identifier below). The role answers «in what capacity,» not «which specific person.»
The same person might hold multiple roles across different decisions; the same role might be held by different people over time. Recording the role, separately from the individual, keeps the organizational context legible even when personnel change.
authority.id
A declared identifier of who made the decision. Verified in the code: if left blank, this field defaults to the full name of the currently logged-in user, but it can be changed, for the legitimate case of recording a decision made by someone else.
Not every decision recorded in EVIDE was made by the person operating the keyboard. Allowing this field to be edited lets a person accurately record a decision made by a colleague, while still keeping the verified identity of the underlying session separately preserved.
Leave it blank when you, the logged-in user, are the person who made the decision, your name will be used automatically. Fill it in explicitly when you are recording a decision made by a different person.
Do not treat this field as a substitute for, or a modification of, your verified account identity, that identity is preserved separately and automatically, regardless of what you enter here.
Verified in the code: every Human Structured record also carries authority.dapi_number, the DAPI identifier of the actually logged-in, authenticated session, set automatically by the server on every submission and never read from the form. authority.id and authority.dapi_number are deliberately independent: the first is a declared label, the second is the verified identity of the session.
An assistant enters a decision into EVIDE on behalf of a manager who actually made the decision. The assistant keeps their own session authenticated (preserved automatically as authority.dapi_number), but sets authority_id to the manager’s name, accurately reflecting who actually made the decision.
authority_id is not the same thing as an independently verified authority. It is a declared, free-text value that by default matches the logged-in user, but can be overridden.authority.dapi_number.authority.id value has itself been independently verified against an external identity system.authority.verification
A free-text note, optionally declared, that the person recording the decision can use if they intend to link the declared authority to a more specific reference, for example, a DAPI reference for the person or role involved.
It gives the declarant a way to add a more specific reference to the declared authority, without requiring one on every record.
When you want to declare a more specific reference tied to the authority behind this decision, and such a reference is genuinely available.
Do not use this field as if writing it constituted a verification. It is a declared note, not a check performed by EVIDE.
claimed instead of declared in the Identity dimension (the authority_verification_status column and the «Identity» row of the EAR). The Human Structured intake response does not contain this field; on the machine channel the API response includes it, as described in the JSON Schema Reference. This reflects that a reference was declared, not that it was verified. The text appears in the EAR as «Declared authority reference», marked as not verified by EVIDE: the form flags this before submission.A Compliance Officer references an internal DAPI-linked identity as supporting context for the declared authority behind an exception approval, without that reference being independently checked by EVIDE.
claimed status describes what was declared, not what was verified.intervention.classification_status
The classification assigned to this decision is considered settled, with no known ambiguity under the categorization active at the time of closure.
Example: a fraud case reviewed and closed with a clear, uncontested categorization.
Does not mean: that the classification is permanently fixed or immune to future revision, only that no ambiguity was known at closure.
The classification is still tentative or pending refinement, it may change as more information emerges.
Example: an active security investigation where the incident category may be revised as the investigation continues.
Does not mean: that the decision itself is incomplete, the decision can still be finalized while its classification remains provisional.
The classification is assigned despite a known interpretive disagreement, or an overlap with another category.
Example: a decision where two reasonable people could categorize the same case differently, and that disagreement is known at closure.
Does not mean: that the underlying decision is invalid, only that its categorization carries an acknowledged interpretive tension.
It makes the operational quality of a classification observable at the time of deposit, rather than presenting every classification with the same implicit, undeclared confidence.
When you have a view on how settled, provisional or contested the classification assigned to this decision is.
If you have no basis for judging the stability of the classification, it is acceptable to leave this field unset rather than guess, the field is optional, and the system assumes no default state if it is absent.
intervention.classification_context.threshold_status
A relevant decisional threshold is declared to have been met.
Example: an internal policy required a documented justification above a certain risk level, and one was provided.
A relevant decisional threshold is declared not to have been met.
Example: the review found that the conditions required for automatic approval were not satisfied, requiring manual intervention.
A threshold exists in this context, but whether it was met is not known at the time of closure. This is different from not_defined - here, a threshold genuinely applies, but its status could not be established.
No threshold was defined for this case. This is different from unknown - here, there is no applicable threshold to evaluate, rather than an unresolved one.
A threshold, in this context, is a predefined decision criterion, a policy limit, an operational rule, an approval condition, against which a decision may be evaluated. This field declares whether such a threshold was met.
It exposes whether a decisional threshold applied to this case, and if so, its declared status, information that is often implicit and undocumented in ordinary workflow tools.
When a defined threshold, rule or policy condition is relevant to this decision.
When no such threshold genuinely applies to this decision, in that case, not_defined is the accurate value, rather than simply leaving the field blank, if you have reason to actively declare that no threshold applies.
threshold_reference below, if threshold_status is not_defined, the reference is hidden in the form and not submitted; if a value arrives anyway, the request is rejected rather than the value being silently discarded.unknown must not be read as equivalent to «no threshold exists.» That is what not_defined means. unknown means that a threshold applies, but its status could not be established.intervention.classification_context.threshold_reference
A free-text reference to the specific threshold, policy or rule cited in threshold_status above.
It allows the declared threshold to be traced back to something concrete, a policy document, a procedure, an internal control, rather than remaining an unattributed general statement.
When threshold_status is met, not_met, or unknown, and you have a specific reference available, for example, a policy name, a procedure code or a document identifier, given only as illustrative examples compatible with this free-text field.
threshold_status is not_defined, a value in this field is rejected: declaring a reference to a threshold that is declared not to exist would be contradictory. In the form, the field is hidden and not submitted, while the typed value remains in the form if you switch back to a different status. Without a threshold status, the reference provided is retained.intervention.classification_context.threshold_authority.attribution_status
A single, identifiable, attributable authority for the threshold was defined at closure.
Example: a policy explicitly owned by a named committee.
When to use it: when you can point to one clear owner of the threshold.
The threshold is defined across multiple sources, with no single attributable authority.
Example: a rule assembled from several overlapping internal guidance documents with no single owner.
Common misinterpretation: fragmented does not mean that the threshold is invalid, only that its ownership is distributed.
The threshold is present in practice but was not formally defined or attributable at closure.
Example: an unwritten but consistently applied internal norm.
Common misinterpretation: implicit does not mean that informal thresholds are illegitimate, only that they lack formal attribution.
The authority behind the threshold could not be verified at the moment of closure.
Example: the declarant is aware that a threshold exists but cannot identify who owns it.
Threshold attribution describes how clearly the authority or ownership behind the threshold cited above can be traced to a specific person, role or committee, not whether the threshold itself was met.
It becomes relevant precisely when a threshold status of met or not_met is declared, but the ownership of that threshold is not clearly attributable to a single source, making that ambiguity itself part of the preserved record.
When you have a view on how clearly attributable the ownership of the cited threshold is.
intervention.trace.access
A free-text reference that lets someone later locate the broader context of this decision within the system where it actually took place, a ticket, a case ID, an audit trail reference, a log reference or a workflow reference, given as illustrative examples of what is semantically compatible with this field.
A decision rarely exists in isolation from the broader operational trail that surrounded it. This field preserves a pointer to that trail, without EVIDE needing to receive or store the trail itself.
Whenever a specific, locatable reference exists in another system, a ticket number, a case file or an audit log entry, that a later reviewer could use to find more context.
handoff.boundary_readiness.status
No independent gate has assessed this object. The declarant states only that they consider it ready, without an independent check having been performed.
When to choose it: this is the honest default when no separate readiness mechanism was involved, it is not a weaker or lesser value, it is the accurate one for that situation.
An independent readiness gate confirmed stability across what it declares to be complete visibility, with no unresolved signals.
When to choose it: only when a genuinely separate mechanism, distinct from the person or system that made the decision, performed this assessment.
Common misinterpretation: this value cannot be self-declared simply by feeling confident about a decision; it specifically means that an independent gate was involved.
An independent gate confirmed what it was able to observe, but explicitly declares that some relevant signals remain unresolved.
When to choose it: when a readiness gate was involved but its visibility was genuinely incomplete, requires at least one entry in unresolved_signals.
A readiness gate attempted an assessment, but the visibility available to it was insufficient to reach any stability conclusion. The record retains the declared assessment attempt: it does not in itself demonstrate diligence, nor that the process failed.
When to choose it: when a genuine attempt at a readiness assessment could not reach a conclusion due to insufficient visibility, it also requires at least one entry in unresolved_signals.
Boundary readiness declares how ready this decision is considered to be, at the moment of closure, for its evidentiary preservation, specifically, whether an independent readiness gate assessed it, and what that gate could and could not observe. It is a declaration about the state of readiness, not a property that EVIDE itself measures or confirms.
Different decisions carry very different degrees of independently confirmed readiness before they are preserved. Collapsing all of them into a single undifferentiated state would hide exactly the information a later reviewer most needs: was this decision independently checked before being recorded, and if so, how completely?
Always, this field is required for every Human Structured Intake.
Ask yourself concretely: did a mechanism genuinely separate from the decision-maker assess the readiness of this object before it was recorded? If no such mechanism was involved, the honest value is candidate. If one was involved, choose among verified, verified_partial, or unverifiable based on how genuinely complete that mechanism’s visibility was.
Verified in the code: visibility_surface (an internal field in the canonical payload) is derived mechanically from this value through a fixed mapping, it is never a separate choice made by the person filling in the form. This value also governs whether readiness_gate_identifier/readiness_gate_scope_reference and unresolved_signals are required (see Conditional Field Relationships below for the exact rules).
candidate, both readiness_gate_identifier and readiness_gate_scope_reference become required together. If this value is verified_partial or unverifiable, at least one entry in unresolved_signals becomes required. Verified in the code: with candidate, a gate identifier or scope, or any unresolved signals, are rejected; with verified, unresolved signals are rejected. The chosen status is never modified and no submitted data is silently discarded. The declarant’s uncertainties are recorded, with any status, in declared_uncertainties.A technical manager authorizing a production restart after an anomaly might declare verified_partial if an automated diagnostic gate checked most, but not all, of the relevant sensors before the restart, with the specific unchecked sensor recorded as an unresolved signal.
verified in this field is a declared value in the data model, not an assessment performed by EVIDE. EVIDE does not itself check whether the declared gate actually operated, or whether its conclusions were accurate.candidate as a deficient or incomplete submission, it is the structurally accurate value whenever no independent readiness gate was genuinely involved.verified describes what was declared about an upstream or independent gate, it is never EVIDE’s own verification of the decision.handoff.boundary_readiness.readiness_gate.identifier
A gate, in this context, is any mechanism, human or automated, that performed the readiness assessment declared in boundary_readiness_status. This field names or identifies that mechanism.
It allows the declared readiness assessment to be traced back to something specific, rather than remaining an unattributed general statement.
Required whenever boundary_readiness_status is anything other than candidate - that is, whenever an independent gate is being declared.
readiness_gate_scope_reference whenever boundary_readiness_status is not candidate. If either is missing under that condition, the request is rejected. With the candidate status, neither field should be provided: if submitted, the request is rejected, because candidate means that no separate mechanism has been declared.A diagnostic check used before authorizing a production restart, or the internal procedure code governing a compliance exception review.
handoff.boundary_readiness.readiness_gate.scope_reference
While readiness_gate_identifier names the mechanism or control declared to have operated, this field describes the perimeter or scope that mechanism is declared to have covered.
A named gate is more useful when its coverage is also specified, the same diagnostic check might cover an entire facility in one case and a single machine in another.
Required under the same condition as readiness_gate_identifier above: whenever boundary_readiness_status is anything other than candidate.
Distinct from readiness_gate_identifier: one names the mechanism, the other describes what it is declared to have covered.
For a production-line restart, the identifier might name the diagnostic control used, while the scope reference names the specific machine or line that control covered.
handoff.boundary_readiness.unresolved_signals
An unresolved signal is a specific element that remained unresolved or unattributed at the moment the readiness assessment was made, the declared «why» behind a verified_partial or unverifiable result, not just the fact that one occurred.
Without this list, a reader encountering «Boundary readiness: verified_partial» would know that something was incomplete, but not what, often the single most useful piece of information for a later reconstruction of the decision.
Required whenever boundary_readiness_status is verified_partial or unverifiable - at least one entry is required in that case.
Not allowed with candidate (no separate mechanism declared) or with verified (visibility declared complete): if submitted, the request is rejected. The declarant’s personal uncertainties are not signals of a verification mechanism: they belong in declared_uncertainties.
List each specific element that remained unresolved, as concretely as possible, multiple entries are supported when more than one signal remained open.
Directly governed by boundary_readiness_status: required (minimum one entry) when that value is verified_partial or unverifiable.
boundary_readiness_status is verified_partial or unverifiable and this array is empty, the request is rejected; with candidate or verified, a non-empty array is rejected. Empty entries and values that are not text are rejected, not ignored.For the SOC scenario in the Enterprise Use Cases above, when a threat-hunting team separate from the responder examined the endpoints with incomplete visibility (verified_partial), this might read: «Origin of two API requests not yet attributed.» If only the responder carried out the assessment, the same sentence is a declared uncertainty.
declared_uncertainties (additive extension of schema 2.1, declared in extensions)
What the declaring person still considers uncertain at the moment of decision closure. Each entry contains a text (statement, required) and an optional context (context, for example a case reference).
Unresolved signals belong to a verification mechanism separate from the decision-maker. Without this field, a declarant choosing candidate would have had no place, within the hashed record, for what they themselves considered uncertain.
With any readiness value, including candidate and verified, when something remains that you consider uncertain. With verified_partial, the declarant’s uncertainties coexist with the verification mechanism’s signals and remain distinct from them.
One entry for each uncertainty, as concretely as possible. The order in which you declare them is retained. A maximum of 20 entries; text up to 1000 characters, context up to 500.
In the SOC case, the responder authorizes continued operation without any separate mechanism having assessed the conditions (candidate) and declares: «Possible EDR sensor false positive», with context «Case SOC-2026-0412».
The rules below are verified directly against the EVIDE validator - not assumed from field names or general documentation patterns.
| If… | Then… |
|---|---|
threshold_status = not_defined | A value in threshold_reference is rejected. In the form, the field is hidden and not submitted. Without a threshold status, the reference provided is retained. |
boundary_readiness_status = candidate | Gate identifier and scope and unresolved_signals[] are not allowed: if submitted, the request is rejected. |
boundary_readiness_status = verified | unresolved_signals[] is not allowed: if open signals remain, the correct status is verified_partial. |
boundary_readiness_status ≠ candidate | readiness_gate_identifier AND readiness_gate_scope_reference both become required together. If either is missing, the request is rejected. |
boundary_readiness_status ∈ {verified_partial, unverifiable} | unresolved_signals[] becomes required, with a minimum of one entry. An empty array is rejected under this condition. |
An evidence_references[] entry includes a hash_value | hash_scope and hashed_by both become required together for that entry. A partial hash declaration, a value without both of these, is rejected, as is hashed_by or hash_scope without a value. |
declared_uncertainties[] | Independent of readiness: allowed with every status, it neither modifies nor replaces the gate’s signals. |
| An unexpected field, or one of the wrong type | The request is rejected: no submitted field is silently ignored. |
visibility_surface, an internal field inside the canonical payload, is always derived mechanically from boundary_readiness_status through a fixed mapping. It is never a value chosen directly by the person filling in the form.An external evidence reference lets a Human Structured Intake point to an artifact, a document, a screenshot, a log file, a policy, that stays under the organization’s own control, without that artifact ever being uploaded to or held by EVIDE. Verified in the code: this reuses, without modification, the exact evidence_references[] structure already documented in the JSON Schema Reference - not a parallel or Human-specific mechanism.
This matters because it resolves a genuine and common tension: an organization often needs to preserve a verifiable declaration about a decision, while deliberately avoiding handing a sensitive internal document to any external system. External evidence references let both happen at the same time.
POLICY-CREDIT-REV7. The organization keeps that document under its own control. The Human Structured Intake can preserve a declared reference to that policy and, when available, its SHA-256 digest, without EVIDE ever receiving the document itself.artifact_type | Free-text declared type - e.g. screenshot, pdf, policy_document. Not a fixed enumeration and, in the form, never prefilled: if you do not provide it, the EAR shows «(not declared)». |
pointer | An implementation-specific reference to the artifact’s location. EVIDE does not resolve, dereference or fetch it. |
declared_origin | Free-text description of where the artifact comes from. |
declared_relationship | Free-text description of how the artifact relates to this decision. |
declared_description | Free-text human-readable description. |
declared_retention_status | One of persistent_storage, rolling_buffer, unknown. Verified in the code: never set automatically by default, included only if the person actively selects it. |
A reference can optionally carry a declared SHA-256 digest of the artifact, together with who computed it (hashed_by) and the hash scope (hash_scope): one of full_file (entire file), segment (segment), frame (frame), archive (archive). In the form, the scope is not preselected: it must be chosen. See the dedicated SHA-256 section below for exactly what this establishes and what it does not establish.
SHA-256 is a cryptographic hash function: it takes any file and deterministically produces a fixed-length digest such that even a tiny change to the file produces a completely different digest. In an external evidence reference, a declared SHA-256 digest allows a later party holding the actual artifact to check whether it matches what was declared at the time of intake.
Verified in the code: hash validation on this channel is purely syntactic, the server confirms that the declared value is exactly 64 hexadecimal characters, nothing more. EVIDE does not receive the artifact and cannot check whether the declared digest actually corresponds to it. The declared value is preserved exactly as entered, never normalized to uppercase or lowercase.
The distinction that matters is between five separate things: the external artifact itself (never received by EVIDE); the declared reference to it; the declared digest (optional, checked only syntactically); a later hash comparison that someone else may perform; and EVIDE’s preservation of the declaration - the only one of these five things that EVIDE actually performs.
The general principle: structured preservation of the declaration and its associated declared evidentiary context, exactly as submitted, canonicalized and hashed in a deterministic, tamper-evident way.
Returning to the Compliance scenario from the Enterprise Use Cases: a Compliance Officer, Jane Whitfield, is asked to approve a temporary exception to the standard vendor onboarding checklist, under a documented business justification and a defined review point.
She closes the decision at 16:00 UTC, but records the intake later, at 18:03 UTC, after finishing a call. She declares her role as Compliance Officer, references the internal policy that governs exceptions of this kind, and notes that an internal review board, a mechanism independent of her own decision, confirmed the relevant conditions with complete visibility before she made her decision. She references the internal ticket tracking this exception, and separately declares a policy document with its SHA-256 digest, kept internally rather than uploaded to EVIDE.
Every value below is chosen to be internally consistent and compatible with the conditional rules documented above, fields are included because they fit this specific case, not simply to demonstrate that every field exists.
This is the canonical payload EVIDE constructs from the example above, verified field by field against the EVIDE validator. Fields generated automatically by the server (never read from the form) are marked accordingly.
"evide_schema": "2.1", "source_system": "evide_web_human", // server-generated, fixed for this channel "source_reference": "web:1042:a3f91c2d", // server-generated "source_timestamp_utc": "2026-04-07T18:03:00Z", // server-generated -- when EVIDE processed this intake "decision": { "type": "Compliance exception approval", "status": "finalized", // always fixed on this channel "closure_timestamp_utc": "2026-04-07T16:00:00Z", // declared -- see Decision Closure above "summary": "Temporary exception to vendor onboarding checklist" // copied from the common field "title" }, "authority": { "id": "Jane Whitfield", // declared (defaults to the logged-in user if blank) "role": "Compliance Officer", "dapi_number": "0000000000", // always server-generated from the session, never from the form "verification": "Linked to internal DAPI reference for the review board" }, "intervention": { "classification_status": "stable", "classification_context": { "threshold_status": "met", "threshold_reference": "Internal policy COM-EXC-04, section 3.2", "threshold_authority": { "attribution_status": "attributed" } }, "trace": { "access": "Ticket COMP-2026-0417" } }, "handoff": { "reconstruction_independence": "declared", // fixed "submission_status": "not_submitted", // fixed at the time of intake "acceptance_status": "not_claimed", // fixed at the time of intake "boundary_readiness": { "status": "verified", "visibility_surface": "declared_complete", // derived mechanically from the status above "readiness_gate": { "identifier": "COM-EXC-04 review board", "scope_reference": "Vendor onboarding exceptions, Q2 2026" } } }, "extensions": ["evidence_references"], "evidence_references": [ { "artifact_type": "policy_document", "pointer": "internal://policies/vendor-onboarding-v3", "declared_origin": "internal policy repository", "declared_description": "Policy document supporting this exception approval", "hash": { "algorithm": "SHA-256", "value": "cabbcf9f83fb765cd82569e477b6b3abf51d81eba9b3b92f8e9904be98a1a2f8" }, "hash_scope": "full_file", "hashed_by": "internal document management system" } ]
unresolved_signals is correctly absent here, it is required only when boundary_readiness.status is verified_partial or unverifiable, not verified as in this example.Answers to the questions DPOs, lawyers, auditors and managers ask most often before using Human Structured Intake. Each answer links to the section of this page that covers it in more detail.
An HSI record can only be deposited from an EVIDE account held by the holder of a DAPI identity certification, who signs in with one-time access codes. The depositor’s DAPI number is part of the record and is therefore covered by its hash; in the EAR it appears as “Deposited by (DAPI)”.
The decision-maker and role named in the declaration are declared values: they may or may not be the depositor, and EVIDE does not verify them. Attribution concerns the deposit to the account: it does not prove who physically wrote the text and does not establish legal responsibility.
More: Authority Identifier
No. The time of receipt is recorded by the EVIDE server, in UTC, when the deposit arrives, and it is shown in the EAR: it is recorded separately and is not part of the content on which the record’s hash is computed. It is distinct from the closure time of the decision, which is declared by the depositor. It is not a qualified electronic time stamp under the eIDAS Regulation.
On request, the FEDIS receives a qualified electronic time stamp. Under Article 41 of the eIDAS Regulation, a qualified time stamp enjoys the presumption of the accuracy of the date and time it indicates and of the integrity of the data to which they are bound: here, the FEDIS document, which attests the state of the record at that date. The time stamp does not prove when the declared decision was taken.
It keeps the complete declaration: the completed fields, the references given and the depositor’s DAPI number. The record’s SHA-256 hash is computed on the declaration in canonical form: anyone holding the record can recompute it and check that nothing has changed.
The EAR (Evidentiary Artifact Record) is the readable document that shows the record, the time of receipt and the hashes, and it has a hash of its own.
More: SHA-256
Through an external reference: the identifier or location of the document and, optionally, its SHA-256 hash, stating who computed it and over which portion. The document is not sent to EVIDE, which does not acquire it and does not verify the declared hash.
If the document is produced later, its SHA-256 is recomputed: if it matches the deposited one, the digital content is identical to the content referenced at the time of deposit. A match concerns the content: on its own it does not prove where the file came from or that it was authentic at its origin.
If a file is instead deposited in EVIDE through a separate deposit, EVIDE computes its hash. The file is deleted at the end of the applicable retention period, while the hash remains in the record.
No. A correction or an update is a new deposit linked to the previous one. Both are kept, each with its own time of receipt and its own hash, and the sequence remains visible.
The time elapsed between the declared closure and the deposit also remains visible: EVIDE neither hides it nor justifies it.
The record contains what is declared in the fields (decision, declared closure time, decision-maker and role, classification, thresholds, uncertainties, references), including the depositor’s DAPI number: the record’s SHA-256 hash is computed on this content. The EVIDE identifier, the time of receipt and the computed evidentiary profile are recorded by EVIDE separately, outside the record’s hash, and are shown in the EAR.
The external documents referenced are not acquired: registers, documents, logs, tickets and communications stay in the organization’s systems. For a data breach, for example, one can deposit the decision to notify or not to notify, with its grounds and uncertainties, referring to the entry in the breach register by its identifier and hash, without including the data subjects’ data.
Before the free-text fields, the app and the web interface remind users not to enter health data, unnecessary third-party data or irrelevant confidential information.
The record is not public: it is accessible to the account that deposited it and, if it was deposited for an organization, to the authorized people of that organization. The servers are provided by Aruba S.p.A., in Italy, and the platform can only be reached over encrypted connections.
Informatica in Azienda does not sell, transfer or license the content of deposits and does not use it to train artificial intelligence models; the platform does not send it to external artificial intelligence services. Retention and deletion follow the Privacy Policy: deletion is requested in writing and each request is assessed individually.
The underlying principle remains data minimization: the safest way to protect confidential information is not to deposit it, and to refer to it by its hash.
No. EVIDE keeps the position the invited person declares on a FAST REVIEW representation of the record. EVIDE records the creation of the invitation, not its delivery: the link is shared by the person who invites.
The events the platform observes, such as the opening of the link, do not prove that a page was displayed, read or understood, and acknowledging the information notice is not consent. The position I ACKNOWLEDGE WITHOUT EXPRESSING AGREEMENT (PRENDO ATTO SENZA ESPRIMERE ACCORDO) is likewise a declaration by the person who makes it, not technical proof of reading. The response does not concern the external documents cited in the record, which the invited person does not receive.
A declared acknowledgement is not technical proof of reading.
A FAST REVIEW: a representation of the record generated by the platform from a fixed template, without artificial intelligence, frozen at the time of the invitation with its identifier and its hash. The invited person does not receive the complete record or the external documents it cites, and the response refers to that specific representation.
The FAST REVIEW text is currently in Italian; the positions and the information notices are available in five languages.
The person who responded always sees their own responses; the depositor sees them as long as they have access to the record. For an organization’s record, the person who invites can choose, when creating the invitation, that the response is also visible to the organization’s current members; former members do not see it. The invited person reads this indication before depositing, and the visibility stays the same for their later responses on the same representation.
A response can be included in a FEDIS-LINKED delivered to third parties only if it was deposited after the respondent was shown the FEDIS-LINKED information notice. When objections or personal data are sensitive, it is advisable to choose the most restricted visibility and to write only what is needed.
The invitation expires after 30 days. Expiry, revocation and suspension prevent new responses, but they do not change or delete those already deposited.
An invitation without a response is not a fifth position and does not amount to agreement: it does not say whether the invitation was received, read or ignored. In the FEDIS-LINKED, invitations without a deposit appear only as counts, without names or content.
Yes, by depositing a new response that declares whether it replaces or supplements the previous one. The previous response is kept with its own time of receipt and its own hash, and the depositor of the record cannot change or delete it.
The summary shows each person’s current position: it is an overview of individual declarations, not a vote or a measure of consensus.
The EAR documents the deposit of an HSI record; the EAR-LINKED documents a single linked response and is generated for every response. Both are produced by the platform.
The FEDIS-LINKED is instead a professional document issued on request: only the depositor of the record can request it, with the same procedure and the same economic conditions as any FEDIS. It documents the record and the includable responses present in its snapshot; when it is generated the platform recomputes the hashes and, if one does not match, the document is not generated. A forensic expert examines it and signs it with a qualified electronic signature, and the document receives a qualified time stamp.
More: FEDIS-LINKED
No. HSI documents what was declared, from which account, when EVIDE received it and that the record has not been altered. It does not establish whether the decision was correct, lawful or appropriate, and it does not verify the truth of what was declared.
The assessment of the merits, and the weight to give the record as evidence, rest with whoever has to decide: supervisory authority, court, internal bodies.
Yes. Screenshots of the EVIDE Mobile app are in the section at the top of this page. For a live demonstration, write to app@certifywebcontent.com.
Human Structured Intake exists so that a declared human decision, and its declared evidentiary context, can survive outside the system that produced them, independently retrievable, deterministically hashed, and time-anchored.