UK Regulator Sets Out Guidelines for Responsible Development of AI Foundation Models

The Competition and Markets Authority (CMA) in the UK has recently published a report outlining principles designed to guide the responsible development and use of Foundation Models (FMs) in the field of artificial intelligence (AI). The CMA’s recommendations seek to strike a balance between fostering innovation, ensuring consumer protection, and maintaining healthy competition in the rapidly evolving AI landscape.

Foundation Models, such as ChatGPT and Office 365 Copilot, are versatile AI systems with broad capabilities adaptable for various specific applications. Their potential for driving innovation and economic growth is substantial. These models have the power to transform various industries, offering new products, easier access to information, scientific breakthroughs, and potentially lower prices, thereby benefiting both individuals and businesses.

The CMA’s report highlights the advantages of a well-functioning FM market, emphasizing the positive outcomes it can bring. These include improved consumer products and services, access to valuable information, and increased competition that can boost overall economic growth.

However, the CMA also warns of potential drawbacks if competition falters or developers neglect consumer protection regulations. These issues could lead to consumers being exposed to misleading information and AI-enabled fraud. Furthermore, the concentration of power in the hands of a few dominant firms could result in limited choices and higher prices for consumers.

The proposed principles presented by the CMA aim to provide a roadmap for the ongoing development and use of FMs. These principles draw upon lessons from the evolution of other technology markets and are as follows:

  1. Accountability: FM developers and deployers must be accountable for the outcomes they provide to consumers.
  2. Access: Unrestricted access to essential inputs is crucial for FM development.
  3. Diversity: Encouraging a variety of business models, both open and closed, promotes competition.
  4. Choice: Businesses should have the freedom to decide how to utilize FMs.
  5. Flexibility: The ability to switch between or use multiple FMs according to need should be upheld.
  6. Fair Dealing: Anti-competitive practices such as self-preferencing, tying, or bundling must be avoided.
  7. Transparency: Consumers and businesses should receive information about the risks and limitations of FM-generated content to make informed choices.

Over the coming months, the CMA plans to engage with various stakeholders to refine these principles further. This broad program of engagement will facilitate collaborative efforts to shape AI markets in ways that foster competition and safeguard consumers.

Sarah Cardell, CEO of the CMA, stressed the importance of proactively guiding AI development to ensure it benefits society as a whole. She acknowledged the potential of AI to enhance productivity and simplify daily tasks but warned against the risk of AI concentration in the hands of a few, potentially undermining consumer trust and market fairness.

The CMA’s principles come at a time when many governments and regulators worldwide are exploring ways to regulate AI, especially in the context of foundational models. The European Union and China have already implemented or proposed regulations, while the United States is actively working on its approach to AI governance. The CMA’s guidance aligns with the international trend of ensuring AI benefits society while managing potential risks.