How Synthetic Outlaws Are Scored

Each domain is scored 0–10 across four dimensions of governance failure. Scores reflect published case law, regulatory findings, and investigative journalism as of March 2026.

Bypass

1–2

Rule intent and outcome are fully aligned. AI stays within the scope the rule intended.

3–4

Minor drift between rule text and AI behavior. Edge cases only; the rule still functions as intended most of the time.

5–6

AI routinely achieves outcomes the rule was designed to prevent, through formally compliant means.

7–8

AI systematically produces prohibited outcomes. Compliance and violation are structurally indistinguishable.

9–10

The rule is functionally inoperative against this AI deployment. Compliant violation is the designed use case.

Diffusion

1–2

Clear single accountable party. Direct causal chain from decision to outcome.

3–4

Accountability is shared but traceable. A responsible party can be identified with effort.

5–6

Multiple parties share responsibility. Attribution is contested and outcome-dependent.

7–8

The accountability chain breaks under scrutiny. No single party can be held liable for the harm.

9–10

Accountability is architecturally impossible. Diffusion across actors and layers is total by design.

Capture

1–2

Regulator is fully independent. Adversarial relationship with regulated entities is structurally intact.

3–4

Structural distance is maintained. Some revolving door or information asymmetry present but bounded.

5–6

Regulator depends on regulated entities for technical expertise. Independence is functionally compromised.

7–8

Regulatory capture is structural, not incidental. Industry shapes the standards it is then measured against.

9–10

Regulator is materially dependent on or effectively controlled by the regulated entity.

Governance Gap

1–2

Existing legal framework directly and adequately addresses the AI behavior in question.

3–4

Framework needs modest extension or interpretation to cover AI. Gap is visible but bridgeable.

5–6

Framework requires significant reinterpretation. Gaps are substantial and actively contested in courts or agencies.

7–8

No applicable framework exists. The AI operates in a legal vacuum or under rules designed for a different technology.

9–10

The framework predates the relevant technology by decades and is inapplicable at every level. No legislative or regulatory fix is imminent.