This paper proposes a direction for the Isle of Man, rather than announcing adopted government policy. It connects the Island’s experience in registration and professional services to a practical question: can a small jurisdiction make delegated AI activity easier to trust, govern and trade, while ensuring that residents participate in its benefits?
1 The strategic opportunity
The Island should compete on the quality of its institutions and services. A company using AI agents needs identifiable responsibility, lawful access to data, reliable operational limits, evidence of performance and a route to remedy. These requirements create possible demand for legal, assurance, cybersecurity, insurance and administrative expertise.
Existing registration businesses offer relevant experience. They identify an asset or entity, record responsible parties and interests, support searches and deal with change over time. That experience is a starting point for a new proposition. It is not evidence that the existing powers of a ship or aircraft registrar already extend to software.
A credible promise would be precise: an identifiable organisation, a defined agent deployment, a verifiable mandate, documented data permissions and current assurance. The Island would need to demonstrate that customers value that combination more than a simpler contractual service.
2 Five connected records
The legal principal identifies the organisation responsible for the activity, its relevant authorisations and accountable contact. Registration alone does not permit every form of regulated business.
The deployment record identifies a running configuration, its supplier, dependencies and material version. A reusable model name is not a complete description of how an agent behaves inside a customer’s systems.
The mandate specifies permitted tasks, data, counterparties, financial limits, approvals, expiry and escalation. Operational controls must enforce it. Publishing a mandate without limiting the agent’s actual credentials supplies little protection.
Data permissions identify the product and lawful licensor, with provenance, permitted purposes, restrictions, update obligations and withdrawal arrangements. Training, retrieval, analysis and redistribution are different uses. A registration entry cannot cure a missing licence or remove privacy obligations.
The assurance record identifies who assessed which configuration, against what tasks and acceptance criteria, for how long. Material changes need reassessment. Incidents, complaints, corrections and expiry belong in the lifecycle record. Sensitive underlying evidence requires controlled access rather than indiscriminate publication.
3 Data assets and the international context
Data can have economic value through its usefulness, the rights governing its use and the services maintaining it. Raw facts, a maintained dataset, a licence, a database investment and a registry entry are different objects. Their legal and financial effects should be explained separately.
The Manx Data Asset Foundation initiative is relevant because it proposes a dedicated governance vehicle and a property-creating registration mechanism. The recovered amended Bill describes that design. Parliamentary proceedings announce Royal Assent in 2026, but the exact enacted text, applicable commencement and final operative instruments remain unresolved in this research. An October event report targets launch in Q1 2027. This paper consequently does not represent statutory registration as open today. See the legal-status audit.
The Island is not entering an empty field. China launched a national data-property registration trial in September 2026. Korea explicitly protects qualifying data assets; Japan protects qualifying shared data against unfair appropriation; Vietnam has a Data Law and exchange rules; Europe regulates intermediaries and recognises data-altruism organisations. Each addresses a different part of the problem. These comparisons do not establish an exact foreign equivalent, but they prevent an unsupported claim that nobody else has recognised or registered data-related rights. See the international comparison.
The competitive question is the additional benefit of a Manx arrangement. Does it reduce the time to establish rights? Improve enforceability or accountability? Make permissioned use more dependable? Support a transaction customers would otherwise decline? Registration counts alone cannot answer those questions.
4 Start with customers and bounded work
Three initial markets merit investigation. First, company and registry-service preparation: explaining published requirements, checking missing information and drafting submissions for authorised review. Second, maritime and aviation administration: reconciling records, monitoring expiry and comparing supplier information. Third, governed data products with rights that can be resolved and buyers willing to pay.
These are proposals for controlled pilots. Initial autonomy should exclude safety certification, statutory determinations and unapproved material commitments. A named person should own the workflow and the decision to expand it. A credible comparison should include a non-AI alternative, realistic error conditions and the full costs of integration and oversight.
Large infrastructure investment should follow demonstrated requirements for confidentiality, resilience, latency and economics. The Island need not reproduce every layer of global computation to build a valuable service economy around its use. Local power, connectivity and recovery constraints still matter.
5 Government must retain practical authority
Formal powers can coexist with operational dependence. A public body may retain a contractual right to intervene while lacking the staff, records or replacement options needed to exercise it. This paper treats that as a risk to test, rather than a prediction that delegation necessarily produces capture.
Procurement should identify inspection, override, service recovery and supplier-exit requirements. Run an exercise before describing those requirements as effective. Measure the time and full cost of restoring a defined service, changing a provider and explaining a consequential decision. Public legitimacy depends on accessible complaints and workable remedies as well as technical performance.
An independent public-interest role is distinct from promoting suppliers or issuing commercial credentials. Conflicts, assessment methods and material failures should be visible. Research collaborators and professional advisers do not acquire statutory decision-making powers through their participation in a pilot.
6 Participation belongs in the economic design
Productive improvement and shared prosperity are separate outcomes. A firm can generate more accepted output with fewer paid hours. Lower prices, new tasks, exports and investment may offset some effects, but their scale and timing require evidence. The academic programme examines this conditional transmission rather than assuming either universal displacement or automatic employment recovery.
Residents can benefit through skilled work, new enterprises, useful services and enforceable claims on productive returns. AI access provides capability; transfers provide purchasing power; dividends distribute funded receipts; ownership can carry control and returns alongside losses and financing obligations. None is a complete substitute for the others.
For each pilot, maintain separate accounts for the operator’s margin, the customer’s improvement, public costs and receipts, and residents’ labour and capital income. Avoid counting a time saving and all output generated from that same time as independent gains. Report actual receipts separately from opportunity estimates.
Skills policy should preserve routes into expertise. Automating introductory tasks may improve immediate output while weakening future recruitment and learning. Paid supervised work, apprenticeships and deliberate exposure to exceptions are candidate responses; they need evaluation for additional learning and progression.
7 A staged programme
During the first three months of an authorised programme, establish a sponsor, a capped budget, the lawful procurement route, core legal advice and an independent evaluator. Interview prospective customers and counterparties. Narrow or stop the proposition if they cannot identify a valuable problem.
During months three to six, choose a bounded pilot, publish a voluntary credential and assessment method, and demonstrate mandates, revocation, complaints and supplier exit. Existing entities and contracts can support early work where lawful; a new compulsory register should follow a demonstrated legal gap rather than precede the demand test.
During months six to twelve, compare up to three controlled pilots against their existing processes. Record accepted output, serious errors, full cost, human supervision and actual willingness to pay. Define unacceptable failures in advance.
During months twelve to twenty-four, scale only services with renewals, viable recurring costs and independently assessed reliability. Seek external acceptance for precisely described credentials. Mutual recognition, foreign privacy compliance and sector permissions require their own legal basis.
The timetable and quantities are proposed design choices, not funding commitments or forecasts.
8 What success would mean
Leadership would mean that customers renew, counterparties accept useful credentials, assessed work meets published standards and limits can be enforced. It would also mean that public bodies can replace suppliers and residents retain more useful capability, income and influence.
A small dashboard can track accepted task performance, serious errors, cost per accepted task, paying and renewing customers, time to verify authority, recovery results, local progression and net public cost. A global number-one claim would require a comparable international dataset that this research does not establish.
The Island’s landscape, heritage and community are part of this strategy. AI should help make the place more capable and rewarding to live in, visit and work with. The durable proposition is dependable institutions supporting useful enterprise, with credible participation in the benefits.
Sources and further reading
- Developed registry and agentic enterprise strategy.
- Registry catalogue and dated international standing.
- International data-asset comparison with original-law references.
- Data Asset Foundations explanation.
- Manx legal-status audit.
- The Agentic Economy academic manuscript.
- Tynwald constitutional guide.
This working paper synthesises the project’s research and proposals. AI assistance was used in drafting and development. It is not a peer-reviewed empirical study, an adopted national strategy or legal advice.