The UK's Competition and Markets Authority (CMA) has published a review into artificial intelligence (AI) foundation models, warning of the potential risks associated with AI, and proposing seven principles to guide their responsible development and use. The report was released on Monday.
Foundation models are AI systems trained on large, unlabeled datasets. They form the basis for large language models such as OpenAI’s GPT-4 and Google’s PaLM, which are used in generative AI applications like ChatGPT. They can be employed for a variety of tasks, from translating text to analyzing medical images.
The CMA's proposed principles include accountability, access, diversity, choice, flexibility, fair dealing, and transparency. These principles aim to ensure that AI foundation model developers and deployers are held accountable for outputs provided to consumers. They also promote ongoing access to key inputs without unnecessary restrictions, sustained diversity of business models, sufficient choice for businesses in how they use foundation models, and transparency about the risks and limitations of foundation model-generated content.
These principles were developed with input from 70 stakeholders, including developers, businesses, consumer and industry organizations, academics, and publicly available information.
While the CMA report acknowledges the benefits that well-developed foundation models can bring to people and businesses, it also warns of potential societal harm if competition is weak or if AI developers fail to comply with consumer protection laws. Examples of such harm could include citizens being exposed to significant levels of false or misleading information or becoming victims of AI-enabled fraud.
The CMA also expressed concerns about potential anticompetitive behavior in the longer term. A small number of firms could dominate the market using foundation models to consolidate their position and deliver overpriced or poor-quality products and services.
Sarah Cardell, CEO of the CMA, highlighted the rapid integration of AI into everyday life for people and businesses. She stressed the potential of the technology to enhance productivity and simplify tasks, but warned against taking a positive future for granted, citing risks of undermining consumer trust or market dominance by a few players.
As part of its engagement program, the CMA plans to continue discussions with a wide range of interested parties, including consumer groups, governments, other regulators, and leading AI foundation model developers such as Anthropic, Google (NASDAQ:GOOGL), Meta (NASDAQ:META), Microsoft (NASDAQ:MSFT), NVIDIA (NASDAQ:NVDA), and OpenAI. The regulator will provide an update on its thinking, including feedback on the proposed principles, in early 2024.
The UK government published a white paper in March outlining its guidelines for the "responsible use" of AI. However, to avoid stifling innovation with heavy-handed legislation, it has delegated AI governance to sectoral regulators who will use existing powers without any new laws.
Gareth Mills, partner at law firm Charles Russell Speechlys, praised the CMA's proactive engagement with the rapidly growing AI sector. He noted that while the principles in the report are broad, they are designed to create a low entry requirement for the sector. This would allow smaller players to compete effectively with more established names while mitigating potential negative impacts of AI technologies on consumers.
This article was generated with the support of AI and reviewed by an editor. For more information see our T&C.