Interoperability can be defined as the ‘the degree to which two products, programs, etc. can be used together, or the quality of being able to be used together’. It refers to the ability of two or more systems or components to exchange information, and to use or analyse the information that has been exchanged. However, technological interoperability requires that the format in which data is generated or stored is standardized and that interconnected systems have the ability to read, analyse or make use of the transmitted information.

There are multiple levels or layers of interoperability –

(a) foundational interoperability, where the data is transmitted by one system and received by another, but not interpreted; (b) structural interoperability, which standardizes the format of receiving data across multiple systems; and

(c) semantic interoperability, which ensures that data flowing between two or more systems can be interpreted at the receiving end (relying on the two underlying layers). This makes it imperative for regulatory regimes to standardise the way in which data is received, processed, or both, by multiple systems, in order to fully harness the power of AI across systems. This may be done through archetypes or templates that are comprehensive and evidence-based, created by domain experts. Interoperability may also be technical or non-technical: the former includes communications, electronics, applications, and multi-database interoperability, while the latter considers organisational, operational, process, cultural and coalition interoperability.

Some of the key benefits of interoperability in AI systems include the ease of processing large amounts of data (or big data), precision of outcomes or outputs so processed and timeliness of processing. In a sector such as healthcare, these elements are crucial to improve the quality of decision-making by healthcare professionals. Interoperability also increases the value of existing networks – something that can increase manifold with AI. There are also social benefits to interoperability, whereby a user expends less effort to use various platforms than if each of them were distinct. However, the transmission and interpretation of data from one system to another (or multiple others) gives rise to concerns relating to privacy and data protection, as rights need to be balanced with the interest or benefits accruing from interoperability. From a competition perspective, interoperability through the imposition of standards by regulatory authorities or incumbents in a sector could potentially keep new entrants out, or exclude innovations that are not based on the industry standard.

The OECD principles on AI have also included interoperability as a principle to increase transparency and promote ethical use of AI going forward; developing industry standards internationally is a recognised priority, and this is seen as an ongoing project at the international level by various participant countries.

In the chapter below, we examine the regulatory regime governing interoperability of AI in various countries, and whether or how such regulations propose to or have navigated some of the issues arising from the move towards increasing interoperability.