Even as panellists defended the importance of partnerships to competition in AI markets, the EC hinted at a deepening scrutiny of deals involving large tech firms
The EC’s workshop was a follow-up to its consultation on competition in virtual worlds and generative AI
On 28 June 2024, the EC hosted a Workshop on Competition in Virtual Worlds and Generative AI. The event followed on from the EC’s consultation on the same subject launched in January 2024 and featured panellists from DG Competition, national competition authorities and a range of tech firms. Discussions broadly focused on the competitive restraints of the key inputs to developing virtual world technologies, including “digital twins” and metaverse applications, as well as generative AI systems. The event was also anchored by a presentation from Benoît Coeuré (President, Autorité de la concurrence (AdC)) on the regulator’s consultation on the competitive landscape of generative AI and remarks from Margrethe Vestager (Executive Vice President, EC) on additional inquiries into big tech firms’ partnerships and dealings in AI markets.
Panellists largely agreed that four key inputs into AI systems acted as potential competitive bottlenecks
Across all of the panels, speakers discussed the familiar tensions of preventing harms and promoting innovation while providing additional context to some of the key competitive concerns raised recently by regulators. These concerns were expressed almost exclusively in reference to four major inputs to developing both virtual worlds technology and generative AI:
Computing power, including hardware and cloud capacity;
Data sets, scaled and specified both for general purpose and more tailored applications;
Technical talent and human resources; and
Capital investments.
Discussing the AdC’s opinion on generative AI, Coeuré also noted that access to the energy necessary to run AI systems may also soon represent another competitive barrier for challenger firms. Representing a high profile European startup, Blanche Savary de Beauregard (General Counsel, Mistral Al) expressed the “virtuous circle” of these resources, noting that in order to attract investment, firms needed to demonstrate computing power and talent, for which investment is needed to acquire in the first place. Sebastiano Toffaletti (Secretary General, European Digital SME Alliance) discussed a similar relationship between these inputs but further noted that the interdependence of these upstream markets could easily be weaponised by dominant players and may be more difficult to target with regulatory intervention.
Open sourcing and other commercial innovations remain a key defence for tech giants facing scrutiny
Unsurprisingly, panellists discussed a range of varying responses to potential limits on competition, including promoting existing commercial practices through to specific regulatory interventions. From a less interventionist perspective, Dominique Costesec (Senior Competition Counsel, Google) argued that open source models as well as innovations in model efficiency would continue to ease potential bottlenecks across all key AI inputs. These points received pushback, however, as Stefan Wagenpfeil (Professor, PFH Private University of Applied Science) noted the value of open source models to competitive markets is dependent on the open sourcing of weights and parameters for those models, which is relatively uncommon at present. Wagenpfeil and other panellists instead proposed immediate action from regulators to address anti-competitive pressures, including mandating interoperability, restricting the use of non-compete agreements for employees, labelling AI-generated content and more specifically defining non-compliant conduct in existing frameworks such as the Digital Markets Act.
Mistral AI outlined the competitive value of AI partnerships amid global regulatory scepticism
As a common target of regulatory scrutiny at the moment, partnerships among AI firms was also a key topic of conversation across the whole of the event. In line with findings from the UK’s Competition and Markets Authority (CMA), Savary de Beauregard similarly noted that partnerships among AI firms can be an efficient route to gaining a foothold in markets for AI inputs when executed correctly, as in the case of Mistral AI and its partnership with Microsoft. That argument did not appear entirely convincing to European regulators, however, as Margrethe Vestager restated the EC’s interests in its ongoing inquiry into the partnership between Microsoft and OpenAI. Vestager also commented on additional attention being paid to a deal struck between Google and Samsung for AI-integration in the device market. As global regulators continue to launch investigations into AI partnerships with big tech, it appears issues of exclusivity and corporate control in agreements will be the core questions for identifying potential mergers.
