Getting ahead of AI: Canadian and U.S. regulators don't want to repeat the mistake they made with social media
Competition authorities in both countries are moving early in an attempt to ensure that AI is not dominated by a few corporate giants the way social media is.
A major lesson from the rise of social media is that if competition authorities don’t have a clear understanding of how emerging digital markets work, those markets can quickly become captured by a handful of major firms. Building on that lesson, the Competition Bureau released a consultation paper on March 20 asking for feedback on the competition dynamics in the Canadian AI market.
On a related note, this week the U.S. Department of Justice (DOJ) and the Federal Trade Commission (FTC) reportedly reached an agreement to divide responsibility for investigating the conduct of three leaders in the artificial intelligence (AI) space. The three companies that will be investigated by U.S. anti-trust enforcers are Nvidia, Microsoft, and OpenAI.
This post will look at these initial attempts by Canadian and U.S. competition authorities to ensure that AI markets remain competitive and do not go the way of social media where a small number of global corporations dominate.
Ensuring competition is only one aspect of effectively regulating AI, of course. Canada’s Artificial Intelligence and Data Act (AIDA), which will likely enter the clause by clause phase of its parliamentary journey this Fall, regulates a range of “high impact” use cases (employment, health, law enforcement, content moderation, etc.), details the powers of an AI Commissioner to enforce the Act, and sets out additional rules for generative AI.
AIDA covers some of the same areas covered by AI statutes in other jurisdictions including the U.K. and U.S. states such as California, but most closely resembles the E.U’s AI Act in terms of its breadth and basic structure (i.e. like the EU AI Act, it is essentially “risk-based”).
The EU is a useful point of reference in discussing competition and AI in that it has complementary legislation to its AI Act that governs competition in the the digital sphere (the Digital Markets Act) as well as a basic framework for regulating content moderation (the Digital Services Act).
The Competition Bureau’s paper is quite tentative but starts off with the following distinctions in defining AI markets:
Markets for AI infrastructure: the supply of compute (for instance, processing power, memory, storage and other resources required for the computational success of any program) and data required for the development of AI,
Markets for AI development: the supply of AI technologies such as AI models, algorithms, and architectures, and
Markets for AI deployment: the supply of customer facing AI products or services.
Looking at markets for AI infrastructure, the Bureau appears to be combining high-performance semiconductors (Nvidia) with large cloud computing platforms (Amazon Web Services, Microsoft’s Azure, Google Cloud Platform, etc.).
The market for AI development presumably includes builders of foundation models (Open AI GPT, Google Gemini, Anthropic, etc.).
Finally, customer facing markets probably means things like chat boxes (Microsoft’s Co-pilot, ChatGPT, etc.) as well as any number of other downstream AI products such as those used by banks to measure credit worthiness, resume sorting software, etc.)
The Competition Bureau initiative, with a deadline for submissions of June 29, is a very preliminary effort to understand the evolving AI market in Canada and where competition issues may emerge. That said, it is a welcome effort. The Competition Act reforms contained in Bills C-19, C-56 and C-59 were developed, in part, for the new challenges posed by digital markets such as AI, and it is good to see the Bureau getting to work on the very complex issues involved in maintaining competitiveness in such markets.
On Wednesday of last week, the New York Times reported that the Justice Department (DOJ) and Federal Trade Commission (FTC) were nearing an agreement to divvy up investigations of potential anticompetitive conduct by some of the world’s largest technology companies in the artificial intelligence industry.
As part of the arrangement, the DOJ is poised to investigate Nvidia and its leading position in supplying the high-end semiconductors underpinning AI computing, while the FTC is set to probe whether Microsoft, and its partner OpenAI, have unfair advantages related to the technology used for foundation models.
There is also scrutiny over Microsoft's multibillion-dollar investment in OpenAI and its partnership integrating OpenAI's technology into Microsoft's products. The FTC is investigating if this deal amounted to an undisclosed acquisition that should have undergone anti-trust review.
As part of the agreement, the FTC is retaining its authority over Amazon and the DOJ is retaining its authority over Google. While both of those companies have already been hit with antitrust lawsuits, AI-related investigations into both were held up pending an agreement. With the agreement, those investigations will proceed.
Nvidia’s profits have exploded over the last few years as its chips, traditionally used for computer graphics, have been adapted for the heavy computational workloads of AI. The company is estimated to have as much as 90% of the market for high-end AI chips, which are often difficult to obtain. The French antitrust agency is already investigating Nvidia, according to press reports and last week Nvidia’s market cap hit $3.012 trillion edging past Apple and just a shade lower than Microsoft’s $3.1 trillion.
Nvidia has faced criticism from customers and competitors for its sales tactics, which some say lock them in at the expense of other options, as well as how it bundles essential related software to the chips, according to two people in the industry.
The FTC has also been in the beginning stages of a probe into the cloud computing market, seeking public comments last year. As part of that inquiry, online rivals and others complained about both Amazon and Microsoft, the No. 1 and No. 2 cloud companies respectively, which control more than 55 per cent of the market.
But the FTC and Justice Department aren’t stopping there in exploring competition issues in the AI space. In fact, they are intensifying their inquiries: The agency is already investigating OpenAI’s data collection practices and reviewing the company’s partnerships with A.I. start-ups.
The competition agenda of Canada and the U.S. has diverged somewhat in the last few years as Canada has prioritized a legislative update of its Competition Act (Bills C-19, C-56 and C-59) while the U.S. has engaged in investigative and enforcement action, much of it related to anti-competitive behaviour amongst tech bemouths such as Google, Meta, Microsoft and Apple.
With the passage of Bill C-59 almost certain before the summer recess, Canada will have completed an historic updating of its competition regime giving the Competition Bureau (hopefully) the tools it needs to take on the digital world and, more particularly, the emerging AI sector.
After badly stumbling at the gate, AIDA seems to have gained some traction spurred on by a strong showing by Canadian civil society groups at the Committee stage. It will be fascinating to see where the clause by clause phase goes when parties vote on AIDA amendments in the fall.
But while Canada, unlike the U.S., is fortunate of have at least a shot at passing comprehensive federal legislation regulating AI before its next federal election, it is still light years behind the U.S. in investigating and enforcing competition in the emerging AI sector.
It is good to see that the Competition Bureau has started things rolling with the very preliminary consultation paper discussed above, but if we want to avoid repeating the mistakes made with social media in its nascent stages, the Bureau must pick up the pace fast.