The United Kingdom is poised to introduce new legislation aimed at mitigating the risks associated with artificial intelligence within the coming year, according to Technology Secretary Peter Kyle. Speaking at the Financial Times’ Future of AI Summit, Kyle emphasised the importance of transitioning from voluntary agreements with AI developers to legally binding accords.
Currently governed by a voluntary agreement, entities like OpenAI, Google DeepMind, and Anthropic allow governments to assess their upcoming AI models for risks before public release. However, the forthcoming AI bill seeks to strengthen this framework by making such agreements mandatory for key developers.
The proposed legislation will notably reconfigure the UK’s AI Safety Institute, granting it autonomous status as an arms-length government entity. This move aims to enhance the institute's independence, allowing it to act more effectively in safeguarding citizen interests. Presently, the institute operates as a part of the Department for Science, Innovation and Technology.
At the core of this legislative initiative is the regulation of frontier AI models, such as ChatGPT, which represent advanced systems capable of generating text, images, and videos. This focus reflects the government’s intent to prioritise oversight of some of the most sophisticated and potentially impactful AI technologies.
In addition to regulatory measures, the government is also set to invest in the computing infrastructure necessary to support the development of sovereign AI models by the UK. This pledge follows criticism faced by the government in August for the discontinuation of funding for an "exascale" supercomputer project at Edinburgh University—a project initially promised £800 million by the preceding Conservative administration.
Exascale computing, defined as performing one billion billion operations per second, is seen as pivotal for the broader adoption of AI technologies. Currently, the United States hosts two functional exascale computers, and experts believe China has developed at least one, though it has not been reported in official compute capacity rankings.
Kyle explained that the cancellation of the Edinburgh project resulted from financial constraints inherited from the former government. He asserted that while the government alone cannot finance the estimated £100 billion required for comprehensive compute infrastructure, it will collaborate with private enterprises and investors to accumulate the necessary capital.
During his address, Kyle acknowledged that previous commitments were out of step with the current demands of the large language model (LLM) sector. He pledged that future announcements regarding compute capacity would be thoroughly planned, funded, and executed.
In a related development, Sarah Cardell, Chief Executive of the Competition and Markets Authority (CMA), expressed optimism about the UK’s potential to lead in AI innovation. She highlighted the CMA’s Digital Markets Unit and its planned regulatory approach, suggesting it will be targeted and proportionate, designed not to inhibit but to promote investment in the UK tech sector. Cardell emphasised that the unique regulatory strategy could serve as a platform for growth, positioning the UK as a dynamic player in AI development.
Source: Noah Wire Services