AI and Facial Recognition: Challenges and Opportunities

Wojciech Wiewiórowski

We know that a substantial amount of data is processed to fuel and improve the machine learning algorithms at the heart of Artificial Intelligence (AI). And we are aware of the rapidly increasing precision and capabilities of ubiquitous surveillance equipment.

But what is the EU’s approach to AI and Facial Recognition? On 13 February 2020, the EDPS organised a workshop to discuss this and debate the challenges and opportunities of AI and Facial Recognition applications. The event gathered together more than 50 world-leading researchers, experts and practitioners from academia, regulators, business and civil society, to share their insights and experiences to deepen our understanding of the challenges ahead and the possible policy responses.

The first thematic session of the EDPS workshop concerned the applicability of the General Data Protection Regulation (GDPR) to AI data processing. It is clear that the GDPR is technology neutral, contains broad, flexible and adaptable definitions and does not specifically address AI. Is specific legislation on AI still needed? To answer this question, we would need to agree on how to apply key GDPR concepts, such as data minimisation, accountability and transparency to AI systems. Moreover, we need to fully apply and enforce existing laws in order to identify potential gaps.

Scientific research and innovation was the theme of the second session. Of particular importance in this context were the perceived contrasts between the principle of data minimisation and big data, the challenges of dealing with personal data that is present in unstructured information and the criteria that research activities must meet in order to apply the GDPR’s safeguards and derogations for scientific research. It is clear that the GDPR applies to datasets used for research in Europe, irrespective of the legislation applicable during their collection.

The third thematic session focused on AI and Facial Recognition that are, in principle, complementary technologies, which can also exist without each other. The former enhances the latter, bringing with it new possibilities and risks. Furthermore, Facial Recognition can be applied in a variety of ways. The use of Facial Recognition to identify an individual among many individuals in a public place is far more intrusive than local, one-to-one face authentication to unlock your smartphone. Moreover, once the infrastructure is in place, Facial Recognition technology may easily be used for other purposes (“function creep”). Moreover, poor quality underlying datasets can result in bias or discrimination; correcting such biases is often a task that is outsourced to countries outside the European Union, with fewer worker protections and rights, so the wider human impact also needs to be considered.  The session highlighted the importance of the precautionary principle, which may even justify a ban or temporary freeze on some uses of the technology where its impact on society and the rights and freedoms of individuals is uncertain.

There was a general consensus at the workshop that as AI can have a significant impact both on individuals and on society as a whole, clear red lines on certain extremely high-risk scenarios need to be defined where the essence of rights and freedoms of individuals may be at stake.

While technology cannot solve all societal problems, it should not exacerbate them. Artificial Intelligence should be used to improve our lives, individually and as a society, but such improvement should never come at the cost of our values and fundamental rights. To this end, we have to carefully consider:
•    What problems do we want to solve with technology?
•    Are there problems that cannot be solved by technology?
•    What new problems could new technologies create?

The ideas and insights gathered at the workshop will help us to better plan and focus our future activities on the various technological, legal and regulatory issues related to Artificial Intelligence and Facial Recognition, including our contribution to the EU public consultation on the White Paper on AI and the other strategic documents unveiled by the European Commission on 19 February 2020.

The high-level discussions at the workshop will also contribute to the EDPS Strategy 2020-2024 and our long-term view of global trends, such as the need to ensure the sustainable development of new technologies and the assessment of their potential impact on fundamental rights.