I have been lucky enough in my career to have strategized, written about, shaped and delivered a few regulatory sandboxes in AI, Robotics, IoT and emerging tech in general. As you may guess, I am a big fan of the instrument, both practically and at the policy implementation level.
This is why I warmly welcome the Department for Science, Innovation & Technology’s (DSIT) Call for Evidence on the proposed AI Growth Lab, or to use another more specific name, AI Regulatory Sandboxes.
I believe this initiative could be a crucial next step to secure the UK’s competitive advantage in the burgeoning AI economy.
But implementations details are key and DSIT is right to consult on it.
The Right Initial Questions
The Lab’s core goal is to establish a pioneering, cross-economy sandbox that enables targeted regulatory modifications under robust safeguard, with the ambition to overcome the outdated rules currently holding back AI adoption across critical sectors like healthcare (e.g., autonomous radiology) and planning.
DSIT is asking the fundamental questions necessary for the Lab’s success, particularly regarding the necessary ‘red lines’ (e.g., consumer protection and safety) and the optimal operating model (Centrally Operated vs. Regulator-Operated Labs). This focus on structural integrity ensures the Lab’s design will balance the need for rapid reform with the maintenance of public trust and the UK’s high regulatory standards.
In my experience there is no right or wrong an how to answer these: it depends. But other things should be considered.
Other Critical Areas for Success
While the current questions are excellent, I urge DSIT to ensure the final design is informed by additional critical areas that maximise the sandbox’s impact. Some of the areas are the following.
Explicit vs implicit regulatory innovation?
About two years ago this week I was in Sydney at a United Nations congress on a panel talking about regulatory innovation in worker’s safety. I was of illustrating the learnings from the latest first-of-a-kind regulatory sandbox in safetytech in the UK, when a senior official from a national regulator told me and the audience that they were doing regulatory innovation with tech companies all the time, organically, without a specific label to it.
The lesson for me was that there is much more that meets the eye going on with regulators, especially in the UK. I would be good to know who is doing what, although a a few reviews have been publihsed on regulatory sandboxes, and they are all reccomended readings!
Make them hyper-collaborative!
AI products rarely operate in a vacuum; they integrate into complex systems involving data providers, infrastructure owners, and use case owners, civic society or even end users.
For the AI Growth Lab to enable testing of genuine, end-to-end applications, it must function more like as a collaborative research lab. The Lab must actively facilitate multi-party interactions, bringing the innovators, their required supply chain partners (e.g., a hospital, a logistics firm, and an AI startup), and the regulator into the same process and testing environment.
This is essential for solving the regulatory “chicken-and-egg” problem where partners fear engaging with an unlicensed technology.
Funding and Financial De-risking: Is there a case for ‘Trial Funding’?
The incentives to work with a regulator and regulatory flexibility should suffice. In theory. However experience shows that in some cases this may not be enough and any given design exercise should consider “trial funding” as an option.
The cost of running a live trial, acquiring real-world data, integrating with legacy systems, and dedicating time to regulatory engagemen, can be prohibitive for many innovative SMEs. The Lab could offer small, focused grants or subsidies to cover the direct costs of testing. While regulatory certainty has been shown to boost a firm’s ability to raise private capital, upfront funding is necessary to lower the barriers to entry before success is demonstrated.
Should it be leveraging existing UK innovation infrastructure?
Regulatory innovation in AI is already happening in the UK, and so is a tremendous amount of AI innovation in general, including in the public sector sphere. To move quickly and efficiently, DSIT should weigh the option of embedding the Lab’s operations within or in close partnership with existing, expert UK institutions, rather than building a parallel structure from scratch.
Without intending to name names, both private and quasi-public organisations like the Alan Turing Institute, the AI Incubator, the Catapult Networks (to name a few) already possess the deep technical expertise and convening power needed for effective running and supervision of dedicated sandbox. A partnership model with robust governance could significantly accelerate the Lab’s launch and impact.
What happens afterwards?
Regulatory changes can be complex and time consuming. Take for instance autonomous cars: we still see too little in the UK despite many publicly funded testbeds and trials in the past decade. A truly forward-looking Lab needs a mechanism to put the learnings into practice, to reap the economic and social dividends. Could there be a fast-track route to regulatory changes? A mechanism for scaling up trials? Can government procurement act as an accelerator?
And, guess what, there are many more dimensions to consider before a full design should be launched.
That said, the AI Growth Lab, aka, regulatory sandboxes, has the potential to offer a real opportunity to unlock billions in economic value by making the UK the best place to develop and launch innovative AI applications. Let’s make sure we get the blueprint right.