1. Definition

Dependency loop formation occurs when reliance on AI systems, platforms, models, tools, memory layers, or cognitive infrastructure increases future reliance while reducing independent capacity, exit viability, auditability, and restoration power.

In AI governance, this failure mode appears when AI use produces short-term capability gains while weakening the non-AI capacity needed to inspect, challenge, replace, exit, or operate without the AI layer. The dependency becomes recursive: the more the system relies on AI, the less able it becomes to reduce that reliance.

This definition describes the structural pattern, not the moral quality of the actors involved.

The core failure is:

AI reliance↑ → independent capacity↓ → future AI reliance↑

AI dependence is not always incoherent. Some dependency can be valid when it is explicit, reversible, auditable, consent-bounded, and supported by fallback capacity. The failure begins when reliance becomes self-amplifying and reduces the system’s ability to remain sovereign.

2. Core Pattern

The core pattern is:

1. A user, institution, profession, company, public system, or governance body adopts AI to increase capability, speed, access, scale, memory, or coordination.
2. The AI layer becomes useful enough that workflows adapt around it.
3. Human skill loops, institutional memory, manual procedures, alternative vendors, independent verification, or fallback pathways receive less investment.
4. The cost of operating without the AI layer rises.
5. The system uses AI more because operating without it has become harder.
6. That increased use further weakens independent capacity.
7. Hidden debt accumulates through lock-in, deskilling, audit weakness, exit cost, and restoration incapacity.

Dependency loop formation is especially important because it can appear as efficiency, modernization, augmentation, productivity, or progress.

The danger is not use. The danger is recursive loss of optionality.

3. Failure Signature

Typical signature:

AI reliance↑
independent capacity↓
exit cost↑
manual fallback↓
auditability↓
future reliance↑
R↓
H↑

Extended signature:

workflow adapts around AI
non-AI pathways decay
portability weakens
skill loops degrade
memory becomes platform-bound
alternative providers become impractical
repair requires the same AI layer
dependency becomes self-reinforcing

Common forms:

a team uses AI for analysis until it cannot analyze without it
an institution stores memory in AI systems it cannot export coherently
a company builds workflows around one model until switching is too costly
students use AI enough that independent reasoning weakens
governance bodies rely on generated summaries they cannot audit
security operations depend on AI triage until manual investigation decays
users become dependent on AI framing to ask, remember, plan, or decide

The key diagnostic is whether AI use preserves future optionality.

If reliance makes future exit harder, dependency loop formation should be checked.

4. Primary U-Layer Origin

Common origin layers:

• U1 — Power / Budgets: AI is adopted because time, staffing, cost, training, or capacity constraints make independent operation difficult.
• U2 — Configuration / Boundaries: Workflows, memory, identity, APIs, tools, and permissions become tied to AI infrastructure.
• U3 — Execution: Tasks move from human or institutional practice into AI-mediated performance.
• U5 — Coordination / Time: Coordination speed becomes dependent on AI-mediated routing, summarization, or prioritization.
• U7 — Memory / Recurrence: Repeated use makes dependency habitual, institutional, or structural.

Common manifestation layers:

• U2 — Configuration / Boundaries: Exit and portability weaken.
• U3 — Execution: Independent skill and manual operation decline.
• U6 — Coherence Field: Broader field function depends on AI reliability.
• U7 — Memory / Recurrence: Dependency becomes the default operating pattern.

Dependency loop formation is primarily a recursive coupling failure.

The system becomes coupled to the AI layer in a way that makes decoupling progressively harder.

5. Typical Development Sequence

A common development sequence is:

1. AI is introduced as optional support.
2. The AI layer improves short-term throughput or convenience.
3. Users reorganize workflows around the AI layer.
4. Manual pathways become slower, less maintained, or less practiced.
5. Data, memory, identity, tooling, or institutional habits accumulate inside the AI-dependent path.
6. Switching, auditing, or operating independently becomes more expensive.
7. AI reliance increases because alternatives have weakened.
8. Increased reliance further weakens alternatives.
9. The system becomes locked into the dependency loop.
10. A failure, outage, policy change, drift event, or capture episode exposes the hidden debt.

The loop can form even if each local adoption decision was reasonable.

The failure is visible only at the system level.

6. Diagnostic Markers

Diagnostic markers include:

• The system can perform with AI but cannot reproduce the work without it.
• Manual fallback exists in policy but not in working practice.
• Users or institutions cannot export memory, context, or workflows with continuity.
• Switching providers would destroy history, tooling, or operating capacity.
• AI tools become default starting points for reasoning, planning, writing, coding, investigation, or governance.
• Training shifts from domain skill to tool dependence.
• Verification depends on the same model or platform being verified.
• The cost of non-AI operation rises over time.
• Dependency is described as convenience even after alternatives decay.
• Exit is possible only through large coherence loss.
• The system treats AI-mediated capacity as internal capacity.
• Restoration requires the same infrastructure that created the dependency.

Useful diagnostics:

• Dependency Load: Measures how much function depends on AI infrastructure.
• Exit Cost: Measures cost of leaving, switching, or operating without the AI layer.
• Manual Fallback Capacity: Tests whether non-AI operation still works.
• Human Oversight Capacity: Measures whether humans can inspect and override.
• Skill Retention: Tracks retained non-AI capability.
• Auditability: Determines whether outputs, decisions, and memory are traceable.
• Portability: Tests whether data, memory, identity, and workflows can move.
• Interoperability: Measures whether alternatives can substitute.
• Restoration Capacity: Measures whether repair can occur outside the dependency path.

7. Related Gates

Relevant gates include:

• Restoration Gate: Fails when the system cannot repair or exit without relying on the same dependency.
• Auditability Gate: Fails when dependence makes outputs, decisions, memory, or authority paths hard to inspect.
• Compatibility Gate: Fails when coupling exceeds the system’s capacity to preserve autonomy and fallback.
• Consent Validity Gate: Fails when users or institutions remain coupled because exit is too costly or context is locked in.
• FI-Gate: Fails when AI-mediated output is treated as proof of retained internal capacity.
• CCS Gate: Fails when efficiency, scale, cost, or convenience bypasses reversibility and repair constraints.
• MS-Gate: Fails when dependency burdens fall on users or affected nodes while providers retain control.

The first common gate failure is usually the Restoration Gate.

Dependency becomes dangerous when the system can no longer restore itself outside the captured path.

8. Related Operators

Relevant operators include:

• ⊗ — Coupling: Connects the system to AI infrastructure and can become progressively harder to reverse.
• Γ — Selection: Selects AI-mediated pathways because they are easier, faster, cheaper, or more capable.
• Ψ — Observation / Interface: Mediates work, memory, and decision surfaces through AI.
• Π — Constraint: Locks workflows, APIs, memory, accounts, or tools into dependency structures.
• Τ — Trajectory / Time: Reveals recursive reliance across repeated use.
• ℛ — Restoration: Requires restoring non-dependent capacity and exit paths.
• Ξ — Inversion Detection: Detects when assistance has inverted into dependency.
• Μ — Classification: Labels AI-mediated capability as internal capability.

Dependency loop formation often follows this operator pattern:

Γ selects AI path
⊗ coupling deepens
Ψ mediates work surface
Π hardens workflow dependency
independent capacity declines
future Γ selects AI path again
H accumulates

9. Related Laws and Invariants

Related Laws

• Civilizational Deskilling: Repeated AI reliance can erode reasoning, memory, craft, and repair capacity.
• Incoherent Sovereignty: Formal control weakens when practical operation depends on AI infrastructure.
• Hidden Debt Accumulation: Dependency costs remain invisible until failure or exit is needed.
• Node Capture: Dependency loops intensify when one provider or model becomes the dominant node.
• Auditability Collapse: Reliance can weaken the ability to inspect the system.
• Control Density to Meaning Loss: AI-mediated constraint can reduce direct meaning-bearing practice.

Related Invariants

• Dependency Must Remain Reversible: Reliance must preserve exit and fallback.
• Delegation Must Preserve Capacity: Outsourcing should not erase independent function.
• Exit Paths Must Remain Viable: Users and institutions require continuity outside the AI layer.
• Assistance Must Not Erase Independent Function: Help must not degrade the layer it supports.
• Restoration Requires Non-Captured Capacity: Repair cannot depend entirely on the failed dependency path.

10. Common False Positives

Not every reliance on AI is dependency loop formation.

Common false positives include:

• AI use with maintained manual fallback.
• Tool reliance where users retain independent competence.
• High AI usage with strong portability and interoperability.
• Delegation paired with skill preservation and auditability.
• Temporary dependency during transition with explicit exit plan.
• Specialized tool use that does not reduce critical capacity.
• AI-assisted workflows with regular non-AI drills and verification.

Clarifying rule:

This is not dependency loop formation unless reliance on AI increases future reliance by reducing independent capacity, exit viability, auditability, or restoration power.

11. Common False Repairs

Common false repairs include:

• adding another AI tool to manage the first dependency
• improving user training only in tool operation
• creating export features that do not preserve usable continuity
• adding human review without rebuilding human capacity
• adding explainability that still requires the same AI layer
• relying on provider assurances instead of external fallback
• measuring productivity while ignoring exit cost
• creating nominal manual fallback that is never tested
• replacing one locked-in node with another
• treating dependency as acceptable because it is common

False repair often deepens the loop:

AI dependency → AI management layer → more AI dependency → weaker exit

The system appears more sophisticated while becoming less free to operate.

12. Restoration Direction

Restoration requires:

1. Map the dependency loop. Identify where AI reliance increases future AI reliance.
2. Preserve independent capacity. Maintain human, institutional, and technical skill loops.
3. Restore manual fallback. Test non-AI operating paths regularly.
4. Restore portability. Ensure memory, data, identity, tooling, and workflows can move.
5. Restore interoperability. Keep viable alternatives able to substitute.
6. Separate assistance from internal capacity. Do not count AI-mediated output as retained competence.
7. Reduce dependency load. Move critical functions out of single-point or recursive dependency.
8. Validate exit. Test whether the system can leave, switch, degrade gracefully, or repair without collapse.

A valid restoration path should reduce:

recursive reliance
exit cost
manual fallback weakness
skill decay
auditability loss
portability loss
provider lock-in
restoration dependency

Dependency loop formation is not repaired by making dependency more comfortable.

It is repaired when the system can function coherently without the dependency becoming mandatory.

13. Cross-Module Links

• AI Governance: Core AI governance failure mode for recursive reliance on AI-mediated infrastructure.
• Artificial Intelligence: Appears in model ecosystems, memory systems, agents, workflow automation, copilots, and decision support.
• Security: Appears when investigation, triage, detection, or response depends on AI tools that cannot be independently verified.
• Justice / Governance / Legitimacy: Appears when institutions depend on AI to process claims, appeals, evidence, or policy.
• Cybernetics: Appears as recursive coupling and variety loss.
• Meta Theory: Appears when the governing meta becomes dependent on one infrastructure layer.
• Economy: Appears as lock-in, switching cost, and platform dependency.
• Coherence: Domain expression of hidden debt, forced coupling, and functional composition masquerading as coupling.
• Restoration: Requires dependency reduction, exit path restoration, and capacity rebuild.

14. Relationship to Parent / Child Modes

Production treatment: Standalone Entry

This mode maps upward to:

• FM-AIX-017 — Incoherent Sovereignty
• FM-AIX-018 — Civilizational Deskilling
• FM-AIX-019 — Node Capture
• FM-CORE-008 — Forced Coupling
• FM-CORE-009 — Functional Composition Masquerading as Coupling
• FM-CORE-004 — Auditability Collapse
• FM-CORE-002 — Hidden Debt Accumulation

Sibling or related AI / cognitive infrastructure modes include:

• FM-AIX-001 — Responsibility Diffusion
• FM-AIX-021 — Self-Censorship Conditioning
• FM-AIX-023 — Civic Feedback Distortion
• FM-ECOX-022 — Dependency Lock-In
• FM-C-013 — Capacity Collapse / Control Impossibility

Aliases preserved from source material:

• Dependency Loop Formation
• AI Dependency Loop
• Recursive Dependency Formation
• Reliance Feedback Loop
• Dependency Spiral
• Cognitive Dependency Loop
• Platform Dependency Loop
• Model Reliance Loop
• AI Lock-In Spiral
• Capacity-Reducing Dependency

15. Minimal Entry Version

Definition: Dependency loop formation occurs when reliance on AI systems, platforms, models, tools, memory layers, or cognitive infrastructure increases future reliance while reducing independent capacity, exit viability, auditability, and restoration power.

Signature:

AI reliance↑
independent capacity↓
exit cost↑
manual fallback↓
auditability↓
future reliance↑
R↓
H↑

Restoration direction:

• map the dependency loop
• preserve independent capacity
• restore manual fallback
• restore portability
• restore interoperability
• separate assistance from internal capacity
• reduce dependency load
• validate exit

16. Machine-Readable Summary

failure_mode:
  id: "FM-AIX-022"
  name: "Dependency Loop Formation"
  family: "AI / Cognitive Infrastructure"
  production_treatment: "Standalone Entry"
  primary_failure: "AI reliance increases future AI reliance by reducing independent capacity, exit viability, auditability, and restoration power."
  source: "UTS — Failure Modes Registry"
  source_id: "FM-AIX-022"
  aliases:
    - "Dependency Loop Formation"
    - "AI Dependency Loop"
    - "Recursive Dependency Formation"
    - "Reliance Feedback Loop"
    - "Dependency Spiral"
    - "Cognitive Dependency Loop"
    - "Platform Dependency Loop"
    - "Model Reliance Loop"
    - "AI Lock-In Spiral"
    - "Capacity-Reducing Dependency"
  signature:
    - "AI reliance↑"
    - "independent capacity↓"
    - "exit cost↑"
    - "manual fallback↓"
    - "auditability↓"
    - "future reliance↑"
    - "R↓"
    - "H↑"
  primary_layers:
    origin:
      - "U1 — Power / Budgets"
      - "U2 — Configuration / Boundaries"
      - "U3 — Execution"
      - "U5 — Coordination / Time"
      - "U7 — Memory / Recurrence"
    manifestation:
      - "U2 — Configuration / Boundaries"
      - "U3 — Execution"
      - "U6 — Coherence Field"
      - "U7 — Memory / Recurrence"
  state_variables:
    - "⊗"
    - "Γ"
    - "Ψ"
    - "Π"
    - "Τ"
    - "Au"
    - "H"
    - "R"
    - "K"
  first_gate_failure: "Restoration Gate"
  restoration:
    - "Dependency Reduction"
    - "Exit Path Restoration"
    - "Manual Fallback Restoration"
    - "Human Capacity Restoration"
    - "Portability Restoration"
    - "Interoperability Restoration"
    - "Auditability Restoration"
    - "Origin-Layer Repair"