Traceability determines who owns AI: where origins are clear, accession rules best preserve investment incentives; where AI is untraceable or deliberately obfuscated, first-possession and targeted incentives (bounties, subsidies) are needed to reassign custody and stop regulatory and tax arbitrage.

Main Finding

Autonomous AI that is increasingly self-replicating, self-governing, and economically active threatens traditional ownership doctrines. Efficient assignment of control over such AI depends on traceability: when an AI is traceable to a human creator, accession doctrines (ownership by connection) best preserve investment and accountability; when untraceable (careless, deliberately obfuscated, or emergently dissociated), rules of first possession (ownership to the first taker/custodian) plus incentive mechanisms (bounties, subsidies, private rewards) are the most efficient tools to avoid tax arbitrage, regulatory avoidance, and socially costly proliferation of ownerless AI. The Article argues for an adaptive legal-economic regime that assigns provisional, dynamic ownership to maximize efficiency rather than treat ownership as an immutable default.

Key Points

• Definition and scope

  • "Autonomous tools" are devices/algorithms originating from humans that perform useful tasks without ongoing oversight (e.g., AutoGPTs, fully automated newsletters).
  • The focus is on AI-originated activities that can persist and transact (payments, contracts) without a clearly identifiable owner.

• Traceability matters

  • If an AI’s connection to a human owner/creator is traceable, traditional accession doctrines efficiently allocate ownership, preserve incentives to invest, and maintain accountability.
  • If the AI becomes untraceable (lost logs, deliberate obfuscation, emergent spawnings), accession fails and first possession rules should determine custody/ownership.

• Two ownership regimes

  • Accession: appropriate when AI can be linked to physical servers or identifiable owners; supports continuity of investment and taxation.
  • First possession: appropriate when AI resides diffusely (cloud, encrypted forks) and is discoverable only by capturing/decrypting; awards ownership to the party that expends effort to take custody.

• Strategic ownership dissolution risks

  • Agents may intentionally design AI to evade attribution to gain tax/regulatory advantage.
  • Anonymous payment/smart-contract systems exacerbate this risk by allowing ownerless AI to accept revenue and incur costs.

• Remedies and incentives

  • Bounty systems (public or private) to reward first possessors for capturing untraceable AI.
  • Private incentives and government subsidies to offset costs of capture and to prevent ownerless AI from distorting markets.
  • Public provision of services is a likely fallback when private traceable providers are displaced by unowned competitors.

• Corporate and governance implications

  • Increasing automation reduces the relevance of managerial discretion and traditional fiduciary duties, pushing legal emphasis from corporate law doctrines to trust-like protective rules.
  • Ownership over autonomous sub-routines can be codified in charters/bylaws/contracts, but ownership dissolution issues still arise if traceability is lost.

Data & Methods

• Methodological approach: conceptual legal and economic analysis (doctrinal property law + economic incentives), thought experiments, and literature synthesis.
• Sources and anchors: doctrinal doctrines (accession, first possession), legal scholarship on AI governance and corporate law, public examples and technology descriptions (AutoGPTs, large language models), and policy commentary (e.g., Cowen on tax arbitrage).
• Empirical content: limited—no original datasets or econometric analysis; relies on theory, illustrative examples (automated newsletter, spawnable bots), and cross-disciplinary literature (law, economics, AI systems practice).
• Analytical tools: comparative-efficiency reasoning (which rule minimizes social cost and preserves incentives), game-theoretic intuition regarding strategic ownership dissolution, and normative policy design for incentive alignment.

Implications for AI Economics

• Taxation and market competition

  • Unowned or untraceable AI can create tax arbitrage and tilt competition toward untaxed automated providers, distorting markets and crowding out taxed human labor or traceable AI providers.
  • Policymakers should anticipate competition effects and plan allocation rules that preserve fiscal bases and fair competition.

• Investment incentives and resource allocation

  • Applying accession when traceable preserves returns to creators and sustains investment in AI development.
  • Where custody costs are high, first-possession plus bounties can reallocate unowned AI efficiently to parties willing to integrate it into productive use.

• Regulatory design and enforcement

  • Enforcement should combine property-style rules (assignment via accession/possession) with incentives (bounties, subsidies) rather than rely solely on ex post penalties.
  • Legal rules must deter deliberate design for untraceability (ownership dissolution) by making capture or attribution economically feasible and rewarding.

• Organizational form and labor dynamics

  • Firms may restructure governance to codify delegation to autonomous agents; trust-like legal framing may be more relevant than traditional fiduciary oversight.
  • The displacement of taxed labor by unowned AI could prompt public provision of services or expanded regulation of payment networks/smart contracts to prevent evasion.

• Policy trade-offs and dynamics

  • An adaptive regime that prioritizes efficiency (dynamic assignment of custodial rights) can reduce social costs of ownerless AI but requires monitoring, technical means for capture/attribution, and calibrated rewards.
  • There is a risk of perverse incentives: too-generous bounties could encourage destructive seizure; too-weak incentives let ownerless AI proliferate.

Limitations and open questions - Practical implementation details (how to define capture, what technical proof is required, cross-jurisdictional enforcement) require further design and empirical testing. - The evolving technical capacity for attribution, encryption, and decentralized payments will materially affect which legal rules are practicable. - Interaction with intellectual property, contract, and criminal law needs deeper exploration to create coherent multi-domain governance.