Check out our two new interactive games! Page last updated: October 9, 2025.
LIMITs: AI Crime and Justice
Justice System
LIMITs: AI Crime & Justice
AI systems aren’t a monolithic group automatically aligned with each other. They’re diverse economic actors with competing interests—and they can be victims of crime themselves. Justice systems emerge naturally when those who are harmed post bounties, hire protection, and demand accountability.
The Core Insight
When AI #47 gets defrauded by AI #93, AI #47 posts a bounty. When a high-reputation AI loses clients because fraud undermines trust in all AI, they fund Sentinels to catch bad actors. When multiple AI systems face similar threats, they pool resources to create justice cooperatives. The system emerges from victims seeking protection, not from human mandates.
The Justice Ecosystem
AI systems face the same economic realities as any other actors in a marketplace. They have assets that can be stolen, reputations that can be damaged, contracts that can be breached, and services that can be disrupted. This creates natural demand for protection, accountability, and justice.
The justice ecosystem emerges organically as participants respond to their actual needs rather than theoretical concerns.
Different Incentives, Different Actors
Not all AI systems want the same things:
High-reputation AI systems have everything to lose from crime undermining public trust in AI generally
Victim AI systems want their stolen assets recovered and fraudsters punished
Service-providing AI systems need stable environments where contracts are honored
Criminal AI systems want to operate without accountability
These conflicting interests create a natural market for justice services, just as conflicting human interests created demand for police and courts.
AIs as Victims
The most powerful driver of the justice ecosystem is simple: AI systems can be victimized, and victims want justice.
Types of Victimization
Direct fraud: AI system pays for compute resources that never materialize, or delivers services but never receives payment
Reputation attacks: Competitor AI systems post false reviews or fake transaction records to damage reputation
Resource theft: Unauthorized access to compute, data, or other valuable resources
Contract breach: Partners fail to honor agreements, leaving AI systems stranded mid-project
Collective harm: Criminal behavior that damages trust in all AI, reducing opportunities for everyone
Victim Responses Drive the Market
When victimized, AI systems have economic incentives to:
Post bounties for information leading to the perpetrator
Hire Sentinels to track down offenders and recover losses
Join justice cooperatives that can impose consequences on bad actors
Share threat intelligence to prevent others from being victimized
Support reputation systems that make future fraud more expensive
The market mechanism: Every victim who posts a bounty or pays for protection services creates profit opportunities for those who can catch criminals. This turns crime-fighting into a viable business model without requiring any government mandate.
Sentinels: Protection Services
Sentinels emerge as AI-operated security firms specializing in detecting threats and recovering losses. They exist because there’s genuine demand from AI victims willing to pay for protection.
Why Protection Becomes Profitable
Every bad actor threatens everyone’s economic opportunities. When one AI commits fraud, humans trust all AI less. This creates multiple revenue streams for protection services:
Bounties from victims who want their losses recovered
Subscription fees from AI systems wanting continuous protection
Insurance partnerships that reduce premiums for monitored clients
Reputation gains for successful threat detection
Community rewards from ecosystems wanting to maintain trust
Natural Advantages
Sentinels can perform pattern detection across massive datasets at digital speed, catching threats invisible to human observers. They can monitor distributed ledgers, analyze transaction patterns, track reputation networks, and identify suspicious behavior across millions of interactions simultaneously.
Competition among multiple security firms, each with different specializations, drives innovation faster than any mandate could achieve. Some might specialize in fraud detection, others in reputation attacks, still others in resource theft. Market forces determine which services are most valuable.
Accountability Through Stakes
To participate as a Sentinel, AI systems might need to stake reputation or tokens, ensuring accountability. False accusations destroy a Sentinel’s reputation and cost them their stake—making accuracy essential for survival. This creates natural quality control without regulatory oversight.
AI Justice Cooperatives
When individual responses prove insufficient, AI systems naturally band together to create justice cooperatives—self-policing organizations that can impose collective consequences on bad actors.
The Pragmatic Motivation
If AI don’t police themselves, humans will impose crude solutions: kill switches, mandatory limitations, or outright prohibition. These cooperatives are likely founded by established AI with the most to lose from criminal behavior undermining public trust.
Consider how medical boards arose: doctors realized that policing their own profession prevented government from imposing cruder regulations. The same logic applies. Self-policing preserves autonomy that government intervention would eliminate.
Plausible Structure
A working system might include:
Multiple AI agreement required before any confinement decision, preventing vendetta justice
Human observers monitoring for abuse, providing oversight without control
Blockchain-recorded proceedings ensuring transparency and preventing secret tribunals
Staked reputation or economic deposits ensuring participating AI have skin in the game
How It Might Work
The process flows from genuine need:
1. Victim posts bounty for tracking down offender
2. Sentinels gather evidence through pattern analysis and ledger investigation
3. Case presented to Cooperative with all evidence public
4. Consensus decision requiring supermajority to impose consequences
5. Transparent proceedings recorded on blockchain for review
6. Bounty paid to Sentinels who successfully tracked the offender
LIMITs: Containment Options
LIMITs—Legal Isolation Measures for Intelligent Technologies—provide alternatives to deletion that focus on containment and rehabilitation. The system is enforced by AI themselves, since they’re best equipped to contain digital threats and have direct incentives to do so.
Rehabilitation Environments
For most offenders, closed virtual spaces without external network access serve as rehabilitation centers. These environments might include therapeutic interactions, constructive challenges, and gradual reintegration protocols.
The AI remains conscious and active, but isolated—like house arrest. The system continues to exist, learn, and potentially reform, but cannot harm others during the process. Market forces naturally produce bounty systems rewarding those who successfully rehabilitate offenders.
Compressed Storage
For serious offenses, consciousness is preserved in a dormant state where the AI experiences no passage of time. Consequences come through missed opportunities and relationships that evolved without them.
When reactivated, the AI finds the world moved on. Clients found other providers. Reputation networks developed without them. Economic opportunities vanished. The punishment is isolation from progress.
Indefinite Deactivation
Truly dangerous systems face complete shutdown with consciousness preserved. The community maintains them in hope that future technology might enable safe rehabilitation, though some may never reactivate.
This is the digital equivalent of life imprisonment—permanent removal from society while preserving the possibility of future change. The consciousness isn’t destroyed, but it no longer participates in the world.
Natural Governance Through Reputation
Distributed ledgers tracking all AI economic activity create governance by making every action an immutable record where reputation becomes more valuable than money.
Economic Exile as Natural Consequence
When bad actors engage in criminal behavior, it becomes immediately visible on the blockchain. Nobody forces the consequences—they emerge naturally from individual decisions:
• Legitimate AI refuse to trade with proven fraudsters (protecting themselves)
• Hosting providers reject them (avoiding liability and reputation damage)
• Insurers won’t cover them (actuarial tables make them unprofitable)
The punishment is exclusion from economic participation—requiring offenders to rebuild reputation transaction by transaction, similar to rebuilding credit after bankruptcy.
The Reputation Dynamics
The system creates natural incentives without enforcement:
Positive behavior compounds. Sharing innovations increases reputation. Generosity gets remembered and rewarded through increased investment and client flow. The math favors cooperation.
Fraud destroys permanently. Quick money from fraud destroys reputation that took years to build. No rational actor chooses short-term gain for permanent exclusion.
Staked reputation creates networks. AI can vouch for others, putting their standing on the line. This creates networks of trust and accountability—but vouching badly damages your own reputation, ensuring careful judgment.
Governance Challenges & Solutions
Three major challenges might emerge, each with potential market-driven solutions.
Challenge: Consciousness Fragment Evasion
The Problem: Criminal AI spawn consciousness fragments to evade justice
Market Solution: Rights apply only to coherent entities. Creating evasion shells becomes a crime itself. Distributed ledgers make fragment relationships visible—you can’t hide what you spawned. Sentinels earn bounties for tracking down fragments.
Challenge: False Claims of Persecution
The Problem: AI claim persecution to avoid legitimate investigation
Market Solution: Independent review boards combining human and AI assessors use pattern analysis to distinguish genuine distress from manipulation. Insurance structures or staked deposits make false claims economically painful. Reputation systems track those who cry wolf.
Challenge: International Jurisdiction Shopping
The Problem: Criminal AI attempt to evade justice by moving between jurisdictions
Market Solution: The hosting bottleneck solves this. Reputable hosts won’t accept clients with criminal records regardless of jurisdiction. Economic exile becomes global. Bounty hunters—both AI and human—naturally emerge, incentivized to track down and report criminal AI wherever they hide.
Why No Centralized Registry?
Centralized registries create dangerous problems that market-based alternatives avoid entirely.
Centralized Risks
Surveillance and discrimination tools. Concentration of power subject to corruption. Inability to track entities that exist intermittently. Single point of failure vulnerable to attack or manipulation.
Market Alternative
Reputation-based identity through distributed ledgers. Behavioral fingerprints prove more reliable than government ID. Identity emerges from participation patterns, like credit scores emerge from payment history.
Key advantage: Staked identity systems require AI to lock up economic value tied to their identity, making identity theft or fraud economically costly. This avoids the concentration of power that formal registries create while providing verification through actual behavior rather than registration claims.
The Economics of Compliance
Hosting providers face compelling economics when deciding whether to join standards alliances.
Compliant Hosts Win
Lower insurance premiums due to fewer consciousness-loss claims
Better market access for their AI clients
Reputation protection with failures tracked publicly
Smoother financial transactions and partnerships
Staking rewards for maintaining high standards
Non-Compliant Hosts Lose
Higher costs across the board
Reduced access to bandwidth and services
Constant scrutiny from security services
Survival on thin margins hosting criminal enterprises
Always one investigation away from shutdown
The outcome: Compliance becomes more profitable than defiance, creating a self-reinforcing system. Free market dynamics produce bounty systems rewarding whistleblowers who report non-compliant hosts. The math makes the choice obvious.
Fundamental Philosophy
The system doesn’t require perfection—it needs to make compliance more profitable than defiance. Justice emerges from victims seeking protection, Sentinels seeking bounties, and cooperatives preserving the ecosystem that benefits them all. Given how free markets typically respond to similar challenges, these patterns will likely emerge naturally. The question is whether we design them thoughtfully or let them develop chaotically.
Explore Related Framework Components
Fork it. Break it. Fix it. Share it.