The advent of general-purpose artificial intelligence (GPAI) models poses competition law issues regarding leveraging in downstream markets. There is a need for strict legal standards granting access to competitors to ensure Big Tech's incentives to develop and improve GPAI models.
Last Update: 16.07.24
The emergence of GPAI entails the development of new digital platforms that affect industries across the globe. Large IT companies, known as Big Tech, control key assets, giving these corporations an advantage in these new markets, thus resulting in concentration. Despite Big Tech's current willingness to grant access to third parties, dominant technology companies have the power to deny it. Under current competition law, this could be interpreted as evidence of anticompetitive market leveraging. Although the AI Act is about to be enacted, it has a narrow scope focused on the security and reliability of AI systems. Neither the AI Act nor the DMA directly constrains leveraging GPAI models in downstream markets.
Some proposals to disrupt Big Tech dominance in the GPAI model markets are far-fetched, as they would imply a strong intervention exceeding contestability and fairness goals. More feasible is the idea to regulate access to GPAI models in downstream markets. However, this approach could overlook the relevant meaning of Big Tech companies' development of GPAI models: the existence of fierce competition in this new market. This competition for dominance in the GPAI models market is driven by the incentive to capture the value arising from the breakthrough. Yet, under ex-ante regulation, there is a real risk that these incentives may dissipate, making a strong case for abstaining from regulating mandated access to GPAI models in downstream markets.
Art. 102 TFEU could also deter Big Tech companies from keeping their investments in GPAI development. Despite the strict criteria for granting mandated access developed in the Magill, Bronner and IMS judgments, recent case law has restricted its scope to narrow circumstances while expanding the instances of anti-competitive market leveraging, resulting in finding abuse regarding access to GPAI models on a low standard. Therefore, there is a case for reversing the anti-leveraging enforcement by recovering the strict standard for refusal to deal with access to essential facilities in digital markets.
On the other hand, the benefits of competition in downstream markets and follow-up innovation should not be dismissed. A balance is required. One reason for competition authorities and courts to avoid applying the Magill, Bronner, and IMS criteria in digital markets might be the unsuitability of the indispensability requirement.
While the indispensability requirement element might be relaxed, a strict standard regarding the refusal to deal with access to GPAI models could emphasize the objective justification prong. The landmark 1985 antitrust case Aspen Skiing could be a useful reference. In Aspen, the disputed conduct only had detrimental effects on consumers. Therefore, the feasibility of objective justification and allocating the burden of proof to the plaintiff should ensure Big Tech incentives to improve GPAI models while protecting competition from anticompetitive refusal to deal.
II.1 Technologiegetriebene Märkte
