Identity is not authority
A person's job title, organizational position, or association with an AI system does not establish that the person authorized, directed, or controlled its conduct.
For AI-related litigation and investigations
When AI is involved, reconstructing what happened is only the beginning.
AIEvidenceMap helps legal teams move beyond document review to reconstruct responsibility in complex AI-related matters. Turn fragmented evidence into a structured, defensible account of what the record establishes, where responsibility can be traced, and what remains unresolved.
The problem
AI-related matters present an evidentiary problem that document review alone cannot solve.
When consequential conduct involves AI, legal teams must establish how system activity connects to human decisions, institutional authority, and the surviving record. These relationships often cross organizational and technical boundaries.
Responsibility crosses institutional boundaries
A single matter may involve model developers, technology vendors, deploying organizations, employees, and external service providers. Each may exercise different forms of authority or control. Identifying the participants does not establish who authorized, directed, reviewed, or could have intervened in the consequential conduct.
- Evidence is scattered across custodians and systems
- Contracts, policies, system logs, communications, approval histories, incident reports, and technical documentation may be distributed across multiple organizations. Connecting these records to specific facts, decisions, and actions is essential to reconstructing the matter.
- Authentication does not establish reliability
- A record may establish what a system displayed or produced without establishing whether the machine-generated conclusion was accurate. General system validation does not necessarily establish what happened during a particular historical execution.
- Actual system conduct may not be fully recoverable
- AI systems and autonomous agents may operate through sequences of model interactions, tool calls, intermediate decisions, and human interventions. Execution records may be incomplete or unavailable. The challenge is determining what the surviving evidence, considered together, can establish.
- Document review is not responsibility reconstruction
- Review platforms help legal teams locate, organize, and analyze records. Responsibility reconstruction requires connecting consequential conduct to algorithmic activity, authority, control, governance acts, human decisions, and supporting or contradictory evidence.
How it works
From scattered records to an evidence-based reconstruction.
- 01
Start with the consequential conduct
Define the action, occurrence, decision, or pattern requiring investigation. Identify the people, organizations, AI systems, and known events involved. The investigation begins with the conduct, not with an assumption about who is responsible.
- 02
Connect the evidence to the facts
Add source records and establish which factual propositions they support, document, contradict, or fail to establish. Facts and relationships are recorded once and reused throughout the investigation as new evidence emerges.
- 03
Reconstruct authority, control, and governance acts
Use TRACE—Responsibility, Authority, Control, and Evidence—to examine delegated authority, operational control, human decisions, and evidenced governance acts. Distinguish organizational roles and technical permissions from actions the record actually establishes.
- 04
Assess what the surviving record establishes
AIEvidenceMap applies the Lozen Evidence-State Scale℠ to classify the surviving conduct record and identify material evidentiary gaps. Legal teams review the system-derived assessment and determine its legal significance.
Request Early Access
Built on Evidence-Based Responsibility Reconstruction℠ (EBRR), developed by Lozen Advisory LLC.
Guardrails
Evidence, not assumptions, determines the reconstruction.
AIEvidenceMap preserves the distinctions necessary for defensible investigation. It does not convert technical activity, organizational roles, or incomplete records into unsupported responsibility findings.
Access is not action
Having authority or technical permission to perform an action does not establish that the action occurred. Likewise, the existence of monitoring tools, logs, or oversight policies does not prove that human review or intervention took place.
Missing evidence is not automatically a failed reconstruction
An unavailable technical record is not necessarily a material evidentiary break. AIEvidenceMap evaluates what the surviving record establishes, distinguishes inference from documented fact, and identifies material gaps affecting reconstruction.
Reconstruction is not a liability finding
AIEvidenceMap connects actors to evidenced governance acts and identifies what the record establishes about authority, control, and conduct. It does not determine legal culpability, liability, or the ultimate resolution of a dispute.
Who it’s for
Built for legal teams investigating consequential AI-mediated conduct.
Plaintiff and defense counsel
Litigators
Reconstruct relationships among parties, systems, decisions, and evidence before advancing factual theories, conducting depositions, or challenging an opposing account. Identify supported propositions, contradictions, and material evidentiary gaps.
- Party and institutional relationship mapping
- Fact-to-evidence connections
- Deposition and case theory preparation
Organizations developing and using AI
In-house counsel and compliance
Investigate how authority was delegated, how operational control was exercised, and which governance decisions can be established from organizational records. Distinguish formal policies and oversight requirements from actual conduct.
- Authority and control reconstruction
- Vendor and organizational relationships
- Governance acts and oversight records
Technical and evidentiary reconstruction
Experts and investigators
Examine the relationship between system activity, human intervention, operating conditions, and surviving records. Develop a documented reconstruction that distinguishes established facts, inferences, contradictions, and material evidentiary gaps.
- Algorithmic conduct reconstruction
- Technical and institutional evidence mapping
- Evidence-state assessment and gap identification
Early access
Bring evidence-based reconstruction to your AI investigations.
AIEvidenceMap is being developed for litigation, investigative, and compliance teams handling consequential AI-mediated conduct.
Request early access to learn about the platform, review its investigation workflow, and explore potential evaluation opportunities.