By Vinod Dhall
In April, the government announced the selection of Bengaluru-based Sarvam to develop India’s first home-grown artificial intelligence (AI) large language model (LLM); it is expected to be ready in six months. This is part of the government’s multi-dimensional plan to foster the growth of AI in India, including foundational LLMs, compute capacity, large data sets, and AI talent. Countries across the globe are making similar efforts. For instance, the European Commission (EC) recently put out a report outlining the European Union’s (EU) integrated strategy to establish Europe as a “leading AI continent”. The global race is on, currently led by the US and China.
AI is often referred to as the cornerstone of the fourth Industrial Revolution that is expected to reshape the landscape of industries, finance, governance, and, in fact, society as a whole. AI offers unparalleled efficiency, innovation, and insights. On the other hand, it throws up a worrying number of potential risks such as entrenched bias, ethical concerns, deepfakes, and worsening inequalities. Naturally, therefore, there is a call for regulation, at least in a calibrated manner. One area of heightened regulatory worry is the relationship between AI and competition/antitrust law. This has engaged the attention of competition agencies and governments across countries; the effort has been to identify potential areas of concern and the search for remedial measures.
Recently, leading competition authorities in the US, EU, and UK put out a joint statement on competition in generative AI (GenAI) foundation models and AI products. The joint statement, while acknowledging AI as a technological inflection point with the potential for unparalleled benefits to society, identified areas where growing AI power can hinder competition; it has suggested principles for protecting competition. Individual authorities from these jurisdictions have undertaken studies on the possible damaging effects on competition and put out informative reports. This includes the US Fair Trade Commission’s report on partnerships between cloud service providers and AI developers, the UK Competition and Markets Authority’s report on foundation models and the EC’s report on competition in GenAI and virtual worlds. The broad observations and conclusions from these are along similar lines and are usefully summarised in the EC’s report.
The reports note that the development and deployment of AI call for huge amounts of costly resources — including in the form of supercomputing infrastructure such as graphics processing units and cloud capacity. Another critical input is vast amounts of high-quality data needed for the development and ongoing training of the models. The availability of AI talent and know-how is a major constraint. Access to these inputs requires serious money. The reports observe that such resources are not within the reach of most AI companies. This puts the large and incumbent big tech players in a very advantageous position compared to others. The competition law question is whether the possession of these facilities can be leveraged to thwart competition in both upstream and downstream markets by denying access to certain parties or by giving access to select players either via exclusivity or preferential treatment.
An incumbent digital major may, for example, sew up an exclusive licensing agreement for securing high-quality data from an upstream source, which may make the data unavailable to competitors. Another cited example is of a data player active both in AI development and cloud capacity denying access to its cloud to other developers or deployers or providing access preferentially to certain parties. The EU study expresses similar concerns over downstream markets too, where AI applications are deployed and commercialised, for example, by having exclusivity conditions or through tying/bundling restrictions, self-preferencing, or lock-in strategies.
These competition authorities have, in particular, taken note of acquisitions or investments by large digital players in smaller AI developers. Such arrangements do have the beneficial effect of providing the small players access in both the upstream and downstream markets and facilitating the development as well as distribution of AI systems. These also provide emerging developers much-needed capital and inputs such as large data sets and computing power. Additionally, it provides them distribution channels and access to customers. To the large digital player, it provides intellectual property and skilled manpower. In that sense, this is a win-win situation. However, the downside that competition authorities see with such arrangements is that these may create high levels of concentration and foreclose access of competitors to critical paths to development or deployment.
Authorities have examined (from the merger control perspective) some high-profile investments. For example, in the partnership between Microsoft and Inflection, (like other such deals) Microsoft inter alia obtained a licence for Inflection’s intellectual property rights and acquired practically all its staff including its two founders. The question that arises is whether this deal amounts to a merger under competition law and if so, whether it would adversely affect competition. The EC’s view was that the deal amounted to a structural shift in the market (and hence a merger) because Inflection’s position in the relevant markets for GenAI foundation models and chatbots stood transferred to Microsoft. However, the European Court of Justice has prohibited the EC from examining merger cases which fall below the EU’s merger revenue threshold.
Looking at the rapid emergence of AI in India, supported by the government, the Competition Commission of India (CCI) will sooner or later be confronted with cases in AI markets both on the merger front and antitrust complaints, such as those mentioned above. The regulator will face the issues that have been engaging overseas authorities. This will require a deeper understanding of these markets and a nuanced and balanced approach that avoids an overreach or over-regulation that may thwart innovation or improved services, and, on the other hand, not allow competition in these markets to be chilled. To this end, the CCI has commissioned a study which inter alia aims to help understand key AI systems and markets and emerging and potential competition issues. However, the reports released by other competition authorities can provide valuable insights into emerging competition issues in AI as well.
The writer is senior adviser, Touchstone Partners, and former head, CCI.
Disclaimer: Views expressed are personal and do not reflect the official position or policy of FinancialExpress.com. Reproducing this content without permission is prohibited.