Recently, the UK government announced a new action plan to ensure that regulators and regulation support growth. The paper finds that amongst the many strengths of the current regulatory system, challenges arise particularly for businesses. Reforms aim to simplify processes, reduce uncertainty, and foster a more balanced approach to risk. In response to these challenges and as part of the government’s action plan, there is an increasing focus on AI in driving regulatory efficiency. Some regulators have already committed to incorporating AI.
AI has been steadily shaping our world since Alan Turing articulated it as a concept in the 1950s. Recently, however, Generative AI (Gen AI) has made great leaps forward. The rapid integration of Gen AI tools into workflows and business systems raises a critical question: how do we harness AI’s potential while mitigating its inherent risks, particularly in the realm of law and regulation?
While Gen AI excels in areas such as summarization and simplification, its susceptibility to bias and ‘hallucinations’ poses significant challenges, especially in areas like law and regulation where you have to be right. Regulators face a dual mandate: adapt to the evolving consumption of regulations and strategically integrate AI into workflows. The AI hype cycle might fluctuate but, certainly, AI is not going away.
Over the past decades, while rulemaking has become steadily more digital, the sheer volume of regulations has grown exponentially. Regulators are increasingly adopting technology to improve efficiency and streamline processes. AI looks like the next step, however, crucially it must be applied meaningfully, ensuring that it addresses, rather than exacerbates, existing complexities.
Regulatory foundations: prioritizing data management
Computers have likely been a driver in the expansion of the regulatory footprint in recent decades. Creating a new regulation can be as easy as copying an existing file and tweaking it. This sounds straightforward but it can be prone to human error.
This issue, coupled with the allure of ‘throwing AI at it’, risks further compounding the problem, potentially flooding systems with redundant or flawed regulations.
AI’s dependence on data quality underscores the urgency of structured approaches to data management. The axiom ‘garbage in, garbage out’ is associated with AI. Ultimately, the data that feeds AI systems drives the quality of its outputs. In our view, good data is structured data from which clearly defined relationships can be established. That includes ensuring regulators have the ability to:
- Establish a single source of truth
- Manage regulations as small, reusable snippets, e.g. clauses
- Track relationships between regulations and related materials such as primary legislation, case law, etc.
- Enrich regulation at the e.g. clause level with metadata and tagging
In an AI-driven future, robust data systems are the foundation on which effective systems are built.
Industry perspective
Modern products are increasingly born digital. In other words, they are digitally designed and simulated before any physical manufacturing occurs.
Computer-Aided Design (CAD) systems are used for creating digital models. Computer-Aided Manufacturing (CAM) systems are used to translate designs into production instructions, and Computer-Aided Engineering (CAE) systems are used to simulate performance. As such, many physical buildings, aircraft, medical devices, etc. have ‘digital twins.’
In recent years, the golden thread has emerged as a key concept. It is strongly linked to the construction industry; however, its core principles are universally valuable for ensuring transparency, accountability, and safety across sectors such as:
- Manufacturing: tracking the parts and processes involved in producing complex products such as aircraft or medical devices
- Healthcare: maintaining accurate patient records
- Pharmaceuticals: tracking the development and production of drugs
Essentially, any industry where safety, quality, and regulatory compliance are paramount can benefit from the golden thread approach. The golden thread should be created while the digital twin is in development.
There are numerous benefits to this approach, however, compliance complexities emerge early in the product lifecycle. A part designed in a CAD system may initially meet the regulations applicable at the time, but those regulations evolve.
Ideally, up-to-date compliance information would be accessible within the CAD system and the engineer could select the exact clause or requirement that the part complies with. If something changed, they could see this information within the CAD system also. Currently, this is difficult to achieve as there is no ‘thread’ connecting CAD systems with the regulations.
Leveraging technology: meaningful solutions for regulators and the regulated
Knowledge graphs and Retrieval-Augmented Generation (RAG)
Modern software tools – including AI – offer opportunities to streamline compliance processes and drive progress for regulators and the regulated.
Knowledge graphs and RAG are emerging as key methods in the mitigation of AI-related risks. These techniques are enabled by quality data systems and provide a way of contextualizing data for AI systems. RAG systems work by retrieving data that is relevant to the question and then interpreting that data to provide a response. Knowledge graphs help feed the AI system more accurate, contextual information to give more accurate results.
The opportunity to enhance regulatory processes is considerable. By connecting the regulator’s knowledge graph to that of businesses operating within their industry, regulators can offer rich, up-to-date insights that make initiatives like golden threads more effective.
AI’s strategic role
AI excels at automating repetitive or manual tasks such as summarization and simplification. There are numerous ways this can be incorporated into workflows.
AI also provides a solution to streamlining the regulatory footprint by rapidly identifying duplicate or near-duplicate regulations within vast databases. Future redundancies could also be addressed by enabling real-time checks during the drafting process.
Ultimately the creation and consumption of regulation must always prioritize humans, however, they must become machine-readable and human-understandable. At the heart of this evolution is a shift from thinking about regulations as not only ‘documents’ but also as data.
Regulators face a dual mandate: adapt to the evolving consumption of regulations and strategically integrate AI into workflows.
Beyond the document
The AI revolution demands a rethinking of the pipelines that underpin regulatory systems. This is not a challenge that AI can simply ‘fix’; structured approaches to managing data that do not interrupt the work of regulatory bodies are key to setting organizations up for success in the AI era. Leveraging technologies like knowledge graphs and RAG helps mitigate AI-related risks. This is also an opportunity to foster greater collaboration and a connected ecosystem between regulators and the regulated.
