The regulator launched an investigation into partnerships between tech firms while panellists raised concerns about single points of failure elsewhere in the AI value chain
The FTC unpacked competition concerns throughout the AI stack
On 25 January 2024, the US Federal Trade Commission (FTC) hosted the FTC Tech Summit on Artificial Intelligence (AI) through its Office of Technology, which was formed in 2023. The event featured three expert panels which discussed different points of the AI tech stack:
Computational infrastructure of AI, including computer microprocessors and cloud services;
Data and its use in training AI models; and
AI in consumer applications.
The event marked further progress in the regulator’s ongoing work to monitor the growing use of AI systems and enforce existing competition and consumer protections laws in the field. The FTC has already hosted a roundtable discussion on the implications of generative AI technologies for creative industries and launched an exploratory challenge to study the use of AI technologies in voice cloning.
The infrastructure layer of AI presents a number of single points of failure
To open the summit, Lina Khan (Chair, FTC) likened the current regulatory moment to the decision American regulators made to champion the Boeing and McDonnell Douglas aerospace merger in 1997. By endorsing concentrated production, Khan argued that innovation and quality suffered. Panellists discussing the computational infrastructure of AI predicted a similar story where the benefits of AI would be muted or lost due to a lack of competition in upstream markets. Both Ganesh Sitaraman (Director, Vanderbilt Policy Accelerator for Political Economy and Regulation) and Dave Rauchwerk (Founder, Next Thing Co.) pointed to the dominance of Nvidia in the market for graphics processing units (GPU) which are chips central to the deployment of AI systems. Rauchwerk also mentioned the longtime partnership between Nvidia and TSMC as another example of how the limited competition in both the GPU and semiconductor markets limits AI innovations downstream. Tania Van den Brande (Director of Economics, Ofcom) described the lack of competition in the cloud services market as another, related market impacting competition in AI. Referencing the recent referral of the cloud services market made made by Ofcom, the UK communications regulator, to the Competition and Markets Authority, Van den Brande noted that AI developers looking to enter the market with new and innovative solutions may be unable to scale up and challenge incumbents if they are unable to access affordable cloud computing power given consolidation.
The FTC’s work on AI will continue with an investigation into big tech partnerships
At the beginning of the summit, Khan also announced that the FTC would be launching an inquiry into the investments and partnerships made between generative AI companies and big tech firms. The investigation, launched under the regulator’s so-called “6(b)” authority for general information gathering, will centre on requests for information issued to five companies: Alphabet, Amazon, Anthropic, Microsoft and OpenAI. The FTC requested information on the specific conditions for partnerships, including the operational impacts for the partnering companies, as well as information on the impacts for AI resources and inputs, such as cloud computing capacity or hardware, which could have a significant impact on competition at numerous points along the AI value chain. The FTC’s investigation, in its breadth under the 6(b) authority as well as its focus on numerous partnerships, represents the widest ranging inquiry on competition among AI firms currently underway in the world, even as numerous European regulators also investigate the relationship between OpenAI and Microsoft. As the US has emerged as a global leader not only in AI technology but in the regulation of that technology through President Biden’s Executive Order on AI, the leadership of the FTC and its developing expertise on emerging technologies will be central to continuing to police the harms of AI in the absence of new laws.
