EVIDE for Legal & Evidentiary Use
When an AI system takes part in a decision, the first question is not who is responsible. It is what can actually be demonstrated.
When AI systems participate in decisions, actions, or transactions involving multiple parties - people, organizations, providers, deployers, AI agents, infrastructure, and external systems - establishing legal responsibility isn't always the first problem. Often, the first problem is simpler and comes before it: what can actually be demonstrated about what happened.
That preserved material may subsequently be examined by a competent authority or authorised examiner, according to the authority and procedures applicable to the specific case.
EVIDE does not determine liability. EVIDE preserves what may later be needed to assess it.
From problem to evidentiary action
Every Evidentiary Question in this library follows the same path.
What is the evidentiary problem?
Identify the point where attribution, authority, oversight, or reconstruction becomes difficult.
What would need to be demonstrated?
Separate what is assumed from what would actually need evidentiary support.
What can EVIDE preserve?
Identify the declarations, identity and authority context, continuity, external references, and boundary information EVIDE can actually preserve.
How can it be prepared in practice?
Show the practical EVIDE workflow that can be configured before or during the relevant AI-assisted or agentic activity.
What changes when the issue is later disputed?
How the evidentiary situation differs when contemporaneous structured records exist, instead of relying only on retrospective reconstruction.
Every Evidentiary Question in this library follows this same sequence - from identifying the evidentiary problem to what a later, independent assessment could actually examine. That later assessment belongs to whoever is legitimately entitled to examine the matter - a competent authority or authorised examiner - not to EVIDE itself.
Preserved material, subsequent examination
Preserved evidentiary material does not interpret itself. What EVIDE preserves is bounded, attributable, integrity-protected material - not a conclusion. Once preserved, it becomes available to whoever is legitimately entitled to examine the matter.
That examination can lead to one of three evidentiary outcomes. These are not EVIDE verdicts. They describe possible outcomes of a subsequent examination of the preserved material.
Supported
The preserved material provides evidentiary support for a proposition relevant to the examination.
Not supported
The preserved material does not provide the evidentiary support required for that proposition.
Unresolved
The preserved material is insufficient, conflicting, incomplete, or otherwise does not permit the question to be resolved on that material alone.
UNRESOLVED is a legitimate evidentiary outcome, not a system failure. Preservation does not guarantee that a question can be resolved. It preserves bounded material that may remain available for subsequent examination.
EVIDE preserves and structures evidentiary material, and may expose evidentiary conditions or signals within its defined constructs. EVIDE does not determine legal truth, liability, responsibility, admissibility, evidentiary weight, or the final conclusion of an investigation or proceeding. That evaluation belongs to the competent authority or authorised examiner, according to the applicable authority, procedure and context.
Two ways into this library
Evidentiary Questions
Recurring evidentiary problems raised by AI-assisted and agentic decisions, examined one at a time - what would need to be demonstrated, and what can actually be preserved.
AI Law, Responsibility & Authority
Legislation, regulation, court decisions, and official guidance relevant to authority, oversight, and evidentiary duties around AI systems, organized by jurisdiction.
In progressStart here
Select an evidentiary question from the list below to explore the problem, what would need to be demonstrated, and what EVIDE can preserve.
Does a Human Instruction Authorize Every Action Taken by an AI Agent?
A person gives an AI agent a goal. The agent turns that goal into a sequence of interpretations, tool calls, and interactions with other systems. The instruction is real - but what does it actually demonstrate about everything the agent went on to do?
Does a Human Instruction Authorize Every Action Taken by an AI Agent?
A person gives an AI agent a goal. The agent turns that goal into a sequence of interpretations, tool calls, and interactions with other systems. The instruction is real - but what does it actually demonstrate about everything the agent went on to do?
Does a Human Instruction Authorize Every Action Taken by an AI Agent?
A person gives an AI agent a goal. The agent turns that goal into a sequence of interpretations, tool calls, and interactions with other systems. The instruction is real - but what does it actually demonstrate about everything the agent went on to do?
Law, evidence, and what EVIDE does
A law can define duties, rights, or standards. A court or competent authority can determine responsibility or liability in a specific case. EVIDE preserves structured, independently examinable evidentiary elements that a later assessment may need. EVIDE does not make the legal determination itself, and does not automatically bridge the gap between a declared fact and a legal conclusion about it.
Law defines the standard. EVIDE preserves the material. The competent authority or authorised examiner evaluates what that material supports, does not support, or leaves unresolved. The legal or evidentiary conclusion itself remains external to EVIDE.