How governments can use public procurement to shape the future of AI regulation – and boost innovation and growth

When she was elected European Commission president in 2019, Ursula von der Leyen put forward a coordinated European approach on the human and ethical implications of Artificial Intelligence (AI).

Many scenarios were in the bounds of the imaginable: strong AI legislation similar to the General Data Protection Regulation (GDPR), a ban on facial recognition technologies in public spaces, like the ban in San Francisco, and innovative approaches to AI deployment in the public sector similar to the Canadian Directive on Automated Decision Making.

Now, the public consultation on the European Commission’s White Paper coincides with the COVID-19 pandemic. AI could help in the response effort, as highlighted by UN Global Pulse. AI can be as accurate as human intelligence, save radiologists’ time and diagnose COVID-19 faster and cheaper than standard tests. For example, BenevolentAI, a UK startup, uncovered an already-approved drug as a potential treatment for COVID-19 in just 90 minutes.

These are promising developments, but also raise questions about what accountable and ethical deployment of AI should look like. As this discussion is too important to be left to the industry alone, it’s important for governments to proactively shape the development and deployment of AI technologies.

Adequate and proactive regulation by governments is key. But how can policymakers shape a regulation that mitigates the risks by being restrictive and prescriptive while also supporting openness and innovation? And, even more important, how can regulation also remain flexible as technology matures? These requirements lead to the urgent need to rethink how governments regulate – and specifically, how they use their regulatory power to shape the future of AI.

Agile governance principles can play a role. These principles include collaboration and data-informed decision making as well as flexible and inclusive regulatory concepts, which could help governments capture the dynamics of the industry and the market and stay on top of new developments.

One opportunity for governments to respond to these challenges is by taking an agile approach to public procurement. In the EU alone, more than 250,000 public authorities spend around EUR 2 trillion annually on external services, works and supplies, with a continuously increasing amount spent on software, cloud storage and IT support. By consciously using this commercial power and utilizing public procurement, governments could not only support AI innovation and economic growth but also set standards, with a signaling effect on the market.

To boost innovation and economic growth, governments could consider new approaches to procurement to drive innovation in the market.

For example, flexible routes to market and requests for proposals would enable governments to focus on the challenge rather than on a detailed specification of the technology. AI-specific tender templates, model contract clauses and specific procurement frameworks such as challenge-based procurements (like the UK GovTech catalyst) can support this.

Another idea would be to systematically approach smaller or young companies, which might not have the right incentives or the organizational strength to respond to a public call for proposals. Challenges for them include long tendering processes, late payments and a lack of awareness of opportunities with the public sector. However, as the market evolves quickly, it would be beneficial to include these players in the public discourse. As one route, governments could reach out to start-ups through “show-and-tells” and co-working spaces. This could help ensure that public procurement processes do not hinder young and small innovative firms from obtaining the public sector as a customer.

The proactive and strategic design of public tenders is also an opportunity for governments use public procurement as a lever to foster the development of ethical and accountable AI.

Embedding accountability and ethics in the purchasing cycle would allow governments to use their purchasing power to effectively operationalize these principles in a practical manner. For example, governments can pre-approve suppliers that meet ethical standards similar to the Canadian AI suppliers list or ask for the use of best practices for explainable AI. Evaluation could include a score for interpretability. In addition, AI systems and AI impact assessments, like those described by AINow, should be part of the procurement documentation.

The EU Commission White Paper underlines the importance of responsible and forward-looking adoption of AI by public administrations and mentions dialogues on public procurement. However, it does not explain in detail the goals of transforming public procurement processes or how this transformation would be implemented. When the public consultation ends on 14 June, the EU Commission will finalize the regulatory approach towards AI.

It is desirable, then, that the Commission’s response to the complex demands of AI technologies would not only adopt agile principles, but also kickstart the process of rethinking traditional approaches to public procurement. Today, national governments should move beyond talking about AI in principle, and put into action practical solutions to foster the adoption of AI while mitigating its risks.

The “AI procurement in a box” toolkit by the World Economic Forum’s Artificial Intelligence and Machine Learning Platform provides an overview of best practices for AI procurement and tools that support their implementation by government teams.