Maximising the benefits of AI: Supra-national considerations

May 02, 2018

We recently have had sight of a European Commission draft guidance document on a European approach to artificial intelligence. Entitled “Maximising the Benefits of Artificial Intelligence”, the draft guidance surveys the broad opportunities stemming from the application of AI, the potential opportunities for European industry and the potential socio-economic impact of AI on citizens within the European Union and their fundamental rights.

The Commission advocates an integrated and comprehensive European initiative in relation to AI. This would include:

  • the establishment of a European AI Alliance;
  • building an ethical and legal framework;
  • an ambitious approach to addressing socio-economic issues; and
  • the development of European industrial and technological capacity to allow Europe to take a leading role in the evolution of AI.

Stakeholders in AI and its deployment in Europe will wish to track such developments, as the changing regulatory landscape, strengthened EU investment/research initiatives and the automation of the workforce are all likely to present a range of challenges as well as opportunities for industry.

Legal and ethical considerations for stakeholders in the AI Industry

The Commission’s draft guidance identifies the need for trust and accountability in the development and application of AI.  Specifically:

  • organisations and end consumers alike must be in a position to trust the technology in their interactions with it; and
  • Such interactions ought to occur with the benefit of a predictable legal environment and safeguards to protect fundamental rights and freedoms.

The draft guidance envisages:

  • the launch of a plan to develop a European Charter on AI Ethics by the beginning of 2019;
  • a period of self-regulation in the meantime, with the potential for later review of existing legal frameworks should the development of AI impact upon fundamental rights and freedoms;
  • the establishment of a multi-stakeholder AI Alliance group within the EU (scheduled for July 2018); and
  • a further guidance document on the interpretation of the EU Product Liability Directive in respect of AI (to be published in the second quarter of 2019).

Given the importance of data to the training of AI machine learning systems, it comes as no surprise that, in the draft guidance, the Commission calls for businesses to recognise the importance of the sharing / re-use of privately-held data to assist with machine learning (although no legal measures in that regard are as yet proposed).

In addition to such EU initiatives at the supra-national level, a number of Member States have themselves published (or are in the process of developing) their own strategies to respond to the opportunities and challenges posed by AI. For example, the French government unveiled the Villani Report in March 2018, laying out a strategic economic approach for developing France’s AI ecosystem; and the UK released a similar plan outlining eighteen key recommendations to grow the AI industry in the UK in October 2017.

Such initiatives underscore the need for AI industry stakeholders to be aware of the key challenges and risks posed by AI (as well as the opportunities), including in the areas of safety and liability, privacy, and bias. AI developers, for example, should consider whether their AI technologies ought to be embedded with / developed to reflect (as a matter of compliance by design) legal norms, ethical considerations and respect for fundamental rights. They should also consider the potential for the implementation of accountability measures such as AI certification schemes, algorithmic auditing, and decision logs (showing how and why a decision was made).

As the AI industry develops we envisage that there will be an increasingly sophisticated response from law-makers and regulators.  Retrofitting compliance within IT systems is a costly business (and may be very challenging in AI systems that learn and evolve).  To the extent they can, AI systems ought to be future-proofed by reflecting the regulatory direction of travel.  

The author would like to thank Matthew Gregson, trainee solicitor, for his assistance in preparing this article.