DealMakers AFRICA Q1 2020
Monopoly: AI Edition
by Angelo Tzarevski and Kirsty Gibson
Due to the inherent nature of Artificial Intelligence (AI), AI-powered industries naturally tend towards monopolisation. This is because once the software and algorithms have been developed, AI uses data to continue to learn and find solutions to problems, without (or with minimal) human intervention, and it does so with increased accuracy and efficacy compared to humans. This makes it more difficult for competitors to catch-up. When the first company completes the development of its AI and starts to sell it worldwide before its competitors do, the clock starts ticking and the lapse of time before the next AI emerges on the market determines the competitive scene.
While major technology companies are already using AI technology, its application in traditional industries, such as the healthcare and automotive sectors, is just beginning. For example, in the healthcare industry, AI can already be used to diagnose patients via photographs and can predict whether a patient is at an increased risk for suicide purely from the application of standardised tests. However, more research and development are needed before hospitals and clinics begin using this technology on their patients en masse.
The application of AI in traditional industries such as the healthcare and automotive sectors requires regulation in terms of competition and antitrust law. There is scope for antitrust regulators to step in and regulate economies that contribute to global competition, without stifling innovation and wealth creation. How they will effectively achieve this, however, is still unclear.
Abuse of dominance laws could potentially protect consumers from harmful monopolies. In South Africa, such laws prevent companies that have market power of 35% or more from charging excessive prices, refusing competitors and customers access to essential facilities or scarce products and generally engaging in "exclusionary" acts. The South African Competition Commission has already spent time and resources in understanding digital markets in this respect.
Abuse of dominance laws may also help in closing the data gap in the development of AI. By making the data used by the first AI in a particular market available to subsequent AIs in that same market, when new AIs enter a market, they would theoretically already be on the same level as the original AI. In other words, all AI services in the same industry will enter the market already having the same knowledge as the other AIs available for sale. In the context of the healthcare industry, allowing access to the data analysed from the patients that the first AI has diagnosed would place subsequent similar AIs on the same level playing field.
However, the option of data sharing does come with its own problems, as the privacy of consumers must also be considered when allowing competitors to access consumer data. For AI to succeed, consumers would need to trust that the data collected via the use of a particular service will not be shared or traced back to them.
Recent amendments to the Competition Act 89 of 1998 in South Africa could signal a further starting point in regulating AI. Since the amendments, companies accused of charging excessive prices bear the burden of proving that the prices charged for goods or services are reasonable. Consumers who make use of AI services will thus be protected, as the price charged for services using AI will need to be reasonable, as with every other service. However, this could have the unintended consequence of making it difficult for competitors to enter the market as, if the best (or only) AI in a specific industry is priced reasonably, this is most likely the AI that consumers will use, and there may not be a reason or need for any other AI services in that market.
The list of problems and potential solutions are clearly numerous and varied, however, we will not have all the answers until AI is well established in everyday life and becomes increasingly regulated as a result. Even then, amendments to existing antitrust legislation will need to be made quickly and carefully to ensure that AI services can succeed in a competitive economy without destroying competitive dynamics.
Tzarevski is a Senior Associate, and Gibson, Candidate Attorney, Competition & Antitrust Practice, Baker McKenzie Johannesburg.