Accountability dilution

Class 9 of the Risk Taxonomy 2026 defines the structural risk that arises when AI is embedded in legal decision loops and accountability for outcomes becomes inherently diffuse. In this configuration, responsibility is spread across:

• The individual lawyer using or relying on the AI
• The function (e.g., Legal, Compliance, Risk)
• The external or internal AI vendor
• The board and senior leadership

This diffuseness is itself the risk: even without a specific failure, no single actor has strong enough incentives to invest in the rigor, controls, and discipline that clear, individualized accountability would normally create. Class 9 is therefore the institutional substrate risk class; all other classes in the taxonomy are operational manifestations that sit on top of this substrate.

Three-Level Mitigation Structure

Mitigation of Class 9 operates across three nested levels of accountability:

1. Named-individual level (function-wide posture)
   • Control: AI Task Force Charter (STR-07)
   • Mechanism: Formally designates a single General Counsel (GC) as accountable for the function’s overall AI posture.
   • Mandate: The GC is empowered to enforce the AI governance framework, resolve conflicts, and ensure that structural accountability is not delegated away or diluted.
2. Per-capability level (Tier 3+ capabilities)
   • Control: Delegation-Authority Register Architecture (GOV-14)
   • Mechanism: For every Tier 3+ AI capability (i.e., those with material legal, regulatory, or enterprise impact), a named human owner is recorded as accountable.
   • Scope: Ownership covers design intent, risk assessment, control adherence, and lifecycle decisions (deployment, modification, retirement).
3. Per-decision level (material decisions)
   • Controls:
     • Materiality Calibration Methodology (GOV-16)
     • Evidence Register (GOV-13)
   • Mechanism:
     • GOV-16 defines the materiality thresholds at which Full Human-in-the-Loop (HITL) review is mandatory for AI-influenced decisions.
     • GOV-13 maintains the per-decision audit trail, recording who made or approved the decision, what AI outputs were used, what evidence was considered, and how the final judgment was reached.

Definition of Full Class 9 Exposure

A function is considered to carry full Class 9 exposure if it cannot answer, for any given AI-influenced decision, who is accountable at all three levels:

1. Named-individual level: Who is the GC (or equivalent) accountable for the function’s AI posture under STR-07?
2. Per-capability level: Who is the named human owner of the specific Tier 3+ AI capability involved, as recorded in GOV-14?
3. Per-decision level: Who is accountable for this specific decision, as evidenced by the materiality determination (GOV-16) and the audit trail in the Evidence Register (GOV-13)?

If any of these three questions cannot be answered clearly and promptly, the function is structurally exposed to Class 9 risk, regardless of whether any discrete operational failure has yet occurred.

Positioning Within the Risk Taxonomy

• Class 9: Structural, institutional substrate risk — the background condition that shapes incentives, investment, and discipline around AI in legal decision-making.
• Other classes (1–8, 10+): Operational risk classes — concrete failure modes (e.g., hallucination, bias, leakage, control bypass) whose likelihood and impact are amplified when Class 9 is not mitigated.

In practice, effective Class 9 mitigation means that for any AI-assisted legal decision, the organization can:

• Identify the accountable GC for AI posture (STR-07)
• Identify the accountable owner of the AI capability used (GOV-14)
• Produce the decision-level record showing human accountability, materiality assessment, and evidence trail (GOV-16, GOV-13)

Only when all three are in place does the function move from diffuse, structural exposure to disciplined, traceable accountability for AI in the decision loop.