Shared AI data needs more than a shared platform

An OECD report published on 20 October identifies two bills behind a shared-data project: building the common infrastructure and helping each participant join it. Funding the platform leaves the second bill—and the reason to pay it—unsettled.[1]

This matters for AI projects that need information from several organisations. Imagine a component supplier, a factory and a maintenance company considering a joint failure-prediction model. Each holds a different part of a machine’s history. Connecting the records creates a possible technical opportunity; deciding who contributes work and who receives useful results is a separate commercial conversation. This is an illustrative proposal, not a reported deployment.

A ship needs a reason to slow down

Fintraffic’s Maritime Data Space offers a concrete example of that conversation. The Finnish traffic-management company describes a project that starts with governance, data management and value creation before selecting technology. Its rules include how participants balance benefits and costs. The aim is to improve port visits through information shared across the maritime industry.[2]

One use case combines journey and port information to support arrival planning. Another explores how a vessel can give notice of readiness from sea when its berth is occupied, allowing it to slow down. Fintraffic identifies shipping companies, ports and technology providers as participants; its stated ambition includes better decisions supported by AI and machine learning.[2]

The commercial detail becomes clearer in ESL Shipping’s separate account of Virtual Arrival. A ship can reduce speed to match a revised berthing time while keeping its place in the queue, calculated from its arrival at normal service speed. After arrival, the operator compares actual consumption with the normal-speed case and shares the savings with the customer.[3]

ESL says the approach has operated between Luleå and Oxelösund with SSAB and the Port of Oxelösund since 2021. It reports an average 24% reduction in CO₂ emissions for that application. This is the operator’s reported shipping result; it does not measure a separate contribution from AI.[3]

Make participation worthwhile

For the hypothetical maintenance project, suppose connecting a supplier’s records costs £8,000. If its expected annual benefit is £2,000, first-year participation leaves a £6,000 gap even before ongoing costs. These invented figures show the negotiation plainly: another beneficiary could fund integration, the supplier could receive a more useful service, or the project could begin with a smaller task.

The OECD identifies cost sharing and membership arrangements among viable approaches. It also sees demand for specialised AI training data as a possible source of revenue, while stressing that governed access still has to work at the speed and scale AI development requires.[1]

The maritime example makes the incentive tangible. Keeping a place in the queue allows a ship operator to act on better information without sacrificing its commercial position. Sharing the resulting savings gives another participant a reason to cooperate. A data-sharing agreement becomes valuable when it changes a decision that the participants can afford to make.[3]

Report an error or suggest a correction