CMA publishes final principles on AI Foundation Models

In its initial report, published in September 2023, the CMA proposed principles to guide the development and deployment of FMs towards positive outcomes for competition and consumer protection. You can access our briefing on that report here. In the technical update report, the CMA outlines the key risks to competition in the FM sector, having regard to sector developments over the past 6 months, and confirms its final AI principles and next steps in its AI FM programme of work.

In this briefing, we summarise some key points to note from the technical report and what to expect from the CMA in this area between now and the end of 2024.

Key risks to competition in the FM sector

The CMA identifies the following three main ways in which the development of the FM sector poses a risk to competition:

• firms that control critical inputs for developing FMs may restrict access to them to shield themselves from competition
• powerful incumbents could exploit their positions in consumer or business facing markets to distort choice in FM services and restrict competition in FM deployment
• partnerships involving key players could reinforce or extend existing positions of market power through the value chain

The CMA observes that developers are releasing FMs on a range of platforms, but they're often hosted by the largest technology firms (ie, Amazon, Microsoft, and Google), and FM development still hinges on the availability of AI infrastructure (such as hardware, data, and expertise), to which large technology companies typically have greater access. 

The CMA’s final AI principles

The final principles are broadly the same as the principles proposed in the report published in September 2023, with the exception of one key change – the proposed principle of ‘flexibility’ (ie, flexibility to switch or use multiple FMs according to need) has been replaced with the principle of ‘accountability’ (ie, ensuring that FM developers and deployers take responsibility for FM outputs).

The CMA’s final AI principles include:

• Access – ongoing ready access to inputs
• Diversity – sustained diversity of business models and model types
• Choice – sufficient choice for businesses and consumers so they can decide how to use FMs
• Fair dealing – no anti-competitive conduct
• Transparency – consumers and businesses have the right information about the risks and limitations of FMs
• Accountability – FM developers and deployers are accountable for FM outputs

Impact of the CMA’s AI principles

The CMA’s AI principles aren't legally binding in themselves. However, in the technical report, the CMA “urges” firms to align their business practices with the principles and states that it's ready to use its consumer enforcement powers under the Digital Markets, Competition and Consumers (“DMCC”) Bill (which is expected to come into force in Q3 2024) to “raise standards” in AI-powered markets and, “if necessary, to tackle firms that do not play by the rules through enforcement action”.

As a result of the CMA taking a principles-based approach and given the high-level nature of the principles, there's uncertainty as to what the principles require of businesses from an implementation perspective. What's required will vary on a case-by-case basis. It's important that businesses which develop and/or use FMs are both aware of the principles and the CMA’s willingness to take enforcement action in this area (with additional tools under the DMCC Bill).

What next?

Upcoming publications by the CMA in this area include:

• a paper on AI accelerator chips (the supply of which the CMA identifies as diversifying but still constrained) in due course
• a joint statement with the Information Commissioner’s Office on the interaction between competition, consumer protection and data protection in FMs
• before the end of April, the CMA’s response to the UK Government’s call for regulators to outline their strategic approach to AI

Any businesses developing new AI or digital products should continue to monitor these further developments and should also look out for the launch of the Digital Regulation Cooperation Forum AI and Digital Hub pilot in Spring 2024.