MEF regulatory advisor Serafino Abate shares conclusions from recent  discussion held by Members at a workshop on the possibilities and challenges of artificial intelligence and its regulation.

In recent months, we have heard many calls for regulators, legislators and policy makers to get their act together and come up with innovation friendly, clear rules for AI technology and its applications. Companies such as Google and Microsoft, through their CEOs, have voiced their desire to see some clarity around the way AI technology can be put to use. There is a shared sense that without clear rules, AI technology will be under used, and will not fulfil its potential.

Which, contrary to the mainstream media focus on scaremongering campaigns, is considerable, and represent a once in a generation opportunity to take an innovation leap forward. The current debate around AI regulation seem to have so far produced a lot of noise but little clarity, if we set aside the focus of late on privacy and personal data. What is it exactly that needs regulating? And who should be doing the regulating itself?

Whichever the form chosen, it must be transparent, meaningful and goal-oriented. It must also be open and inclusive – and ensure proper gender, age and racial representation. We want rules that represent our modern, diverse and vibrant societies.“

Some of these issues were raised by MEF members during a recent internal workshop on AI regulation. There was clear agreement that, given the broad, trans-national and trans-sectoral nature of the challenge, industry associations that have the capability to explain the technology, like the MEF, should be an active part of the debate.

Indeed, participants felt that one of the first task at hand is to bridge the current gap in understanding between those tasked with making the rules, and those that work with AI technology. This could be done through different types of engagement between regulators and their industry stakeholders, including through a co-formation process for the regulation itself.

Whichever the form chosen, it must be transparent, meaningful and goal-oriented. It must also be open and inclusive – and ensure proper gender, age and racial representation. We want rules that represent our modern, diverse and vibrant societies. What should be avoided at all cost are command and control, top-down approaches based on limited input from the industry and all the different stakeholders’ groups.

The process to set the rules themselves should consider different aspects. On one side, there’s the treatment and market conditions required for each of the inputs of the AI technology – data, computing power, algorithms. This “input” regulation should take into account all the complementary aspects of those inputs – and advance a common framework, rather than focus on one aspect only. It should consider the implications for liability laws, privacy laws, competition law and trade standards altogether.

On the other hand, there is going to be the need for identifying some do’s and dont’s – rules that help chose between different possible outcomes (“output” regulation), align them with societal priorities, preferences and ethical principles. The final outcome of the regulation-forming process should be a set of actionable rules – not generic statements of abstract principles that everybody can agree with – who doesn’t want a “fair” AI, or “good” outcomes?

In short, when it comes to AI regulation, we urgently need less slogans and more concrete, goal-oriented rules that can set AI-powered progress in full motion, and also an open discussion around how we can use different regulatory approaches, from self-regulation through to co-regulation, to ensure that those that understand AI can play an active part in the process.

Regulatory Advisor, MEF