By Mathias Becuywe, Tatiana Beliaeva, Stephanie Beltran Gautron, Theodore Christakis, Maéva El Bouchikhi, Agnès Guerraz
To contribute to the debate around the use and appropriate regulation of Facial Recognition Technologies, AI Regulation and Skopai have engaged in a unique partnership in order to map the current landscape of start-ups developing facial recognition technologies.
Facial recognition technologies (FRTs) are on the rise, and so is the debate around their use and appropriate regulation.
In Europe, for instance, the European Commission published on April 21, 2021 its draft Artificial Intelligence (AI) regulation which includes several proposals on facial recognition. However, the proposed restrictions are deemed insufficient by numerous civil society actors calling for a global ban on the use of facial recognition in public spaces. The two major data protection authorities in Europe, the European Data Protection Board (EDPB) and the European Data Protection Supervisor (EDPS), are also jointly calling for such a ban.
Against this background, the AI Regulation Chair (MIAI, UGA) and Skopai, a Deep Tech and AI company that specialises in the compilation, analysis and classification of start-ups around the world, have engaged in a unique partnership in order to map the current landscape of start-ups developing facial recognition technologies. In this study, the two research partners have combined their talents in order to examine facial recognition technology-related products developed by start-ups worldwide.
This study therefore explores the main sectors, technologies and business models involved in facial recognition and questions the extent to which these companies take into account data protection rules and human rights. The study ends by focusing on how facial recognition start-ups have used their technology during the Covid-19 pandemic.
A total of 199 start-ups were identified, of which 130 were selected for this study; Skopai’s AI tool was used to collect and process the data from publicly available internet sources. The quality and completeness of the data was verified by innovation experts, and further reviewed and analysed by the AI-Regulation team. The analysis includes a focus on the activities of certain start-ups in relation to specific issues; these are provided as case studies throughout the study.
The study showed that facial recognition start-ups are distributed globally across 36 different countries. Most start-ups’ headquarters are located in the United States (20.9%), India (9.3%), the United Kingdom (7%), and France (6.2%). About half of start-ups (52.2%) focus exclusively on the development of facial recognition technology, but a significant number of them (47.8%) offer a broad portfolio of products. In terms of the technologies offered, face verification is by far the most common functionality (68%). Additionally, more than half of the start-ups (53%) provide face identification functionality, and 47% offer face analysis.
The majority of the start-ups analysed (60%) do not make any public statements about whether or how their products comply with existing safeguards on privacy and data protection. Some start-ups, however, argue that they are concerned about privacy and data protection, as a means of marketing themselves more effectively. They mostly achieve this by referring in general terms to the General Data Protection Regulation (GDPR) or other applicable legal texts. Yet, even among the start-ups that seek to convey a privacy-compliant public image, actual evidence of the legal requirements on data protection being implemented remains sparse.
The study ends with a focus on how facial recognition has been used during the Covid-19 pandemic. Although the available data does not permit us to confirm that the pandemic has been particularly favourable to the proliferation of facial recognition technology, around 20% of the analysed start-ups developed and/or adapted their facial recognition technology in order to offer Covid-19 related services. These include mask detection, using facial recognition to verify the identity of mask wearers, as well as crowd management, to limit overcrowding and verify that social distancing regulations are being respected.
To read or download the full report, please click here.
These statements are attributable to the authors only, and this publication does not necessarily reflect the views of the other members of the AI-Regulation Chair, Skopai, or any other partner organization.
This work has been partially supported by MIAI @ Grenoble Alpes, (ANR-19-P3IA-0003).