The Competition and Markets Authority (CMA) of the United Kingdom, on September 18, published a report including proposed principles to mitigate competition-related concerns in artificial intelligence markets and protect consumer interests. Based on a review of foundational models (FMs) that power AI systems, the CMA highlighted:
“In the long term, a handful of firms could use FMs to gain or entrench positions of market power and fail to offer the best products and services and/or charge high prices.”
Similar to the principles that apply to other technology markets, the CMA, in its review report, has carved out guiding principles for preventing anti-competitive or monopolistic practices in AI markets. The document is meant to guide ongoing development and deployment of FMs, and encourage innovation in order to ensure responsible use of FMs and maintain a ‘positive market outcome’.
Why is a ‘positive market outcome’ important?
The CMA in its review report observed that a positive market outcome is necessary to realize the full potential of Foundational Models (FMs) and develop high-quality models that can be used for a wide range of applications. A positive market outcome can be determined if users, businesses, and the larger economy would gain from the production and appropriate deployment of such FMs. This requires providing suitable market conditions for independent developers to compete with one another, innovate, and allow firms to access such FMs to expand and experiment with their usage.
“In that scenario, firms would be able to experiment with different business models and forms of monetisation, including the supply of FMs on both an open-source and closed-source basis so others can continue to build on existing FM capabilities,” the report states.
Restricted access to inputs, on the other hand, would lead to a concerning market outcome where a handful of firms would control the production and supply chains of leading models, the report says. These firms would strengthen their positions in the market to provide models on a closed-source basis only and to impose unfair prices and terms.
In the context of AI, these market outcomes would be driven by factors such as access to data, computing resources, existing advantages of being a big company or a first-mover, and the existence of competitive open-source models.
Article continues below ⬇, you might also want to read:
- Meta Partners With Microsoft, Unveils Open-Source AI Model ‘Llama 2’ For Research, Commercial Use
- FTC Explains The Competition Concerns Surrounding Generative AI Models: A Summary
- Gatekeepers Under EU DMA: Google, Apple, Meta, Amazon, Microsoft, TikTok
- Google, Microsoft, OpenAI Collaborate To Establish ‘Frontier Model Forum’ For Artificial Intelligence
What are the guiding principles?
- Accountability: The CMA makes it clear that developers and deployers of FMs are accountable for outputs provided to consumers.
- Access: This is to facilitate access to key inputs including computing and data-related resources, expertise, and capital with minimal restrictions. The CMA also emphasizes on preventing successful FM developers from gaining a disproportionate advantage by being the first to develop an FM, having economies of scale, and gaining from feedback loops.
- Diversity: In order to ensure healthy competition, the CMA calls for maintaining “sustained diversity” of business models, including both open and closed. It also states that open-source models prevent barriers to entry in the market and encourage expansion.
- Choice: Industry participants and businesses must have sufficient choice to decide how to develop, use, and deploy FMs, APIs or even to engage in partnerships.
- Flexibility: Further, businesses also must have the flexibility to switch and/or use multiple FMs if needed. This enables interoperability for mixing, matching, or launching multiple FMs. This would allow customers to switch or use multiple services easily and not remain locked in a ‘one provider, one ecosystem’ situation.
- Fair dealing: The fair-dealing principle must alert stakeholders to not indulge in anti-competitive conduct or practices such as self-preferencing, tying, or bundling, that can undermine fair competition.
- Transparency: To further consumer protection goals, users and businesses must have access to information about the risks and limitations of FM-generated content so that they can make informed choices. Developers must also ensure transparency about the processes to help deployers manage responsibilities to their users.
The report also lists factors that would undermine healthy competition in AI markets including, but not limited to:
- Mergers or acquisitions can lessen competition for the development and deployment of FMs.
- Misusing existing positioning in associated markets to block prospective challengers who develop FMs.
- Placing “undue restrictions” that discourage firms from switching between or using multiple FM providers.
- Creating ecosystems, conditions that restrict choice and interoperability.
- Utilizing market power in FM development to engage in anticompetitive conduct such as tying or bundling of products or services.
- Misleading customers with false information to impact their decision-making process.
Why it matters:
UK CMA’s review of foundational models uncovers how rapid development and deployment of FMs can lead to competition concerns in AI and associated technology markets. It also elaborates on factors that would define standards for healthy competition and the actions that would undermine it. Along with concerns about copyright, misinformation, bias-induced discrimination, etc., the rapid development of AI systems has also raised apprehensions about monopolistic practices by Big Tech and leading AI companies.
We are already witnessing collaborations among the Big Tech and companies like OpenAI and Anthropic, for strengthening and expanding their hold over resources for FM research, development, and launch. In June, the United States Federal Trade Commission detailed competition concerns surrounding generative AI. The FTC also mentioned that data, computational resources, and skilled talent are the building blocks of generative AI that can affect competition in the AI market. Unfair methods of competition in the AI industry must be tackled in order to ensure better experimentation, representation of data from varied sources, maintain diversity of approaches towards FMs, and thus ensure equitable deployment.
STAY ON TOP OF TECH NEWS: Our daily newsletter with the top story of the day from MediaNama, delivered to your inbox before 9 AM. Click here to sign up today!