Jonathan Middleton
Director, Financial Services
Artificial Intelligence (AI) is rapidly transforming our world, from voice-activated assistants to autonomous vehicles. However, as the technology becomes more widespread, concerns have arisen about its potential impact on society. In response, the UK government has proposed a risk-based approach to AI regulation, which aims to be flexible and adaptable, keeping pace with technological developments and changes in the AI landscape.
The risk-based approach would mean that the level of regulation applied to a particular AI application would depend on the potential risks it poses. This approach acknowledges the significant potential benefits of AI, such as supporting financial services and creating more efficient businesses, but also recognises the potential risks associated with the technology.
The UK government’s proposed approach to AI regulation involves collaboration between regulators, innovators, and researchers to create a “trustworthy AI ecosystem” that encourages innovation while protecting the public interest. The development of AI requires a multi-disciplinary approach, with input from experts in fields such as ethics, law, and technology.
The government is also considering how to approach transparency about the data and algorithms used in AI systems, with the aim of enabling public trust in the fairness and reliability of AI systems. The establishment of an AI sandbox to enable understanding the latest technology, testing new AI applications in a controlled environment and the creation of a national AI data resource to facilitate research and innovation are other measures proposed by the UK government.
AI Sandboxes provide synthetic data, integrated technology and a Digital Sandbox to test products, industry benchmark datasets and develop AI and ML models that are more accurate and reliable, while minimising risk of bias. Having this innovation visible through a single platform de-risks innovation and gives government a route to understanding the latest technology approaches and encouraging responsible AI innovation.
The Information Commissioner’s Office (ICO) and the Financial Conduct Authority (FCA) have already successfully piloted digital sandboxes in their sectors. The FCA sandbox has worked with over 800 businesses and accelerated their speed to market by an estimated 40% on average. Sandbox participation has also been found to have significant financial benefits, particularly for smaller organisations.
It is clear that AI Sandboxes are a key tool for government and regulators to ensure responsible AI innovation. Early adopters are already using such sandboxes, and given the global opportunities of AI, it is likely many more will emerge around the world.
Some of the key areas for the regulators to focus on:
1. Set expectations for AI life cycle actors to proactively or retrospectively provide information relating to:
(i) the nature and purpose of the AI in question including information relating to any specific outcome.
(ii) the data being used and information relating to training data.
(iii) the logic and process used and where relevant information to support explainability of decision-making and outcomes.
(iiii) accountability for the AI and any specific outcomes.
2. Set explainability requirements, particularly of higher risk systems, to ensure appropriate balance between information needs for regulatory enforcement (for example, around safety) and technical trade-offs with system robustness.
3. Consider the role of available technical standards addressing AI transparency and explainability (such as IEEE 7001, ISO/IEC TS 6254, ISO/IEC 12792) to clarify regulatory guidance and support the implementation of risk treatment measures.
The benefits of a well-designed regulatory framework for AI are significant. A trustworthy AI ecosystem, built on collaboration, transparency, and a risk-based approach, can help to ensure that AI is developed and used in a responsible and ethical manner. This can help to build public trust in AI systems, promote the development of ethical and responsible AI, and ultimately ensure that the potential benefits of AI are realised.
