Facial recognition technology is ubiquitous and has become common in everyday applications, where it is now widely used for verification and identification.
One of these uses is in the news industry, particularly in investigative journalism. This branch of reporting is intrinsically linked to the press’s role as a public watchdog and its efforts to uncover matters of public interest.
However, this use is not exempt from dilemmas and serious risks, ranging from potential violations of data protection to the need to protect the secrecy of investigations or the right to a fair trial.
Natalia Menéndez González from the Center for a Digital Society and Sofia Verza from the Centre for Media Pluralism and Media Freedom, both at the EUI’s Robert Schuman Centre, started reflecting on these issues when BBC Verify, as part of its investigation into the ‘Supernova Music Festival Massacre’ in Israel (October 2023), argued that, by using facial recognition technology (FRT), one of the perpetrators of the murders had been identified as a Palestinian police agent.
In this #EUIResearch interview, they reflect on the human rights implications arising from the use of FRT in investigative journalism, the topic of their recently published article in the Computer Law & Security Review.
Facial recognition technology is becoming increasingly widespread. What exactly is it, and how do these systems work in practice? What are the most common ways they are being used today?
Natalia Menéndez González: FRT compares two or more facial pictures of a person to fulfil one of two goals: identification, which means determining whether the person is part of a database such as employees at a company, students at a university, or crime suspects; and/or verification, meaning checking whether the person is who they claim to be, for example at a border management control. It is important to point out that FRT does not provide a single result. As a probabilistic tool, it provides a percentage, known as confidence score, with the probability of the person being who the system claims.
FRT is currently widely used all around the world both in private and public spaces. Examples of these uses might include: to monitor access at the workplace, schools and universities, banks, gyms, casinos, sport stadiums and conference venues. Further, FRT is highly used by law enforcement authorities all over the world to identify suspects, potential victims, witnesses and missing people.
The use of facial identification techniques in journalism is not entirely new. What makes today’s use of FRT by journalists different from earlier practices of visual identification and verification?
Sofia Verza: Investigative reporters have traditionally used methods like comparing photographs, studying archived images, consulting forensic or technical experts, and checking public databases to identify people. What makes modern facial recognition technology different is not the goal of identification itself, but the speed, scale, and automation involved.
Automated facial recognition systems can quickly compare images against huge image databases, removing many of the practical limits that once slowed investigations. This also makes it easier and cheaper to carry out repeated or large-scale biometric searches. As a result, while the chances of identifying people successfully increase, so do the risks of mistakes, broader surveillance, and the expansion of these practices beyond their original purpose.
You suggest that the use of FRT by investigative journalists, as well as by other civil society actors such as NGOs, points to a growing normalisation of this technology. Even when used with good intentions, how might these practices reproduce broader risks and concerns seen in other contexts?
Natalia Menéndez González: There are multiple risks that the use of FRT poses from a fundamental rights’ perspective, irrespective of who uses the technology. As we mentioned in the article, there are risks when it comes to the rights to privacy, data protection, freedom of expression, and the right to defence to name some. There are also risks when it comes to the rights to fairness and non-discrimination. As it has been widely discussed by the literature, FRT systems do not display the same accuracy rates when it comes to certain demographic groups. Due to this, there is a high risk of false positives that have resulted in the wrong arrests of several people within the US and the UK. In the end, these practices are not new but entail an automatisation and technification of existing racist, colonialist and discriminatory practices. Further, the presence of automation bias (taking as ground truth the results provided by the system) entails a lack of critical thinking and contestation when it comes to processing the outputs of FRT. Continuing with the previous examples, most of the people wrongly arrested because of an incorrect FRT match had an alibi. However, the police did not manage to check that until after they were arrested.
In your paper, you argue that the use of FRT in investigative journalism raises specific and underexplored issues. What makes this use case distinctive, and which fundamental rights and values come into tension here?
Sofia Verza: The literature on FRT use and its legal and democratic implications usually focuses on cases of mass surveillance - such as CCTV or digital cameras placed in public spaces - by public authorities. In this case, instead, we are dealing with private actors, meaning journalists and newsrooms, who however, act in the public interest.
The fundamental rights that need to be balanced, and that we delve into in our article, are the rights to freedom of expression, privacy and data protection, as well as components of the right to a fair trial, including defence rights such as the presumption of innocence.
Also, we conclude that the hard law might not be the ideal regulatory solution to prescribe the procedures to follow when using FRT in journalism. Instead, an ethical use of FRT in journalism could be pursued explicitly in the ethical codes of newsrooms and/or consortia of journalists, and a risk-based assessment of the possible implication of such use should be followed throughout the process leading to the news publication.
Should journalists be treated differently from other private actors when it comes to the use of facial recognition technology?
Natalia Menéndez González: The democratic function that journalists perform cannot be denied. When they are acting in the exercise of their job, they develop a function of democracy watchdogs and, in many cases, accountability seekers. However, this must not entail that they are given carte blanche. As we extensively discussed in our article, the use of FRT by journalists can entail privacy and data protection violations of the people investigated or reported about and it can jeopardise their right to defence in potential future legal proceedings. Finally, the identification of potential criminals, might end up in private revenge or retaliation situations. To avoid this, and to make sure that the fundamental rights of the people under investigation by journalists are respected, a balancing exercise must be done: Is FRT the less restrictive means? The only possible tool to be used? Am I going to blindly follow what the systems tells me? Do I use it as another tool in an extensive investigative toolkit? These are the kinds of questions that journalists and news outlets should reflect upon when considering the use of FRT.
Sofia Verza: For the sake of our article, we interviewed four journalists based in different European countries, who have used FRT for investigative purposes. Theirs is a testimony of highly qualified professionals who follow deontological standards when using such technologies —adopting precautionary measures and double-checking their results multiple times before publication. Presuming that journalists act in the public interest and do follow ethical standards typically associated with the journalistic profession is one of the reasons why journalists should be treated differently than other private actors using FRT. Verifying that this is actually the case, should be done on a case-by-case basis.
Read 'Face the facts': examining the use of facial recognition technology by news outlets through human rights lenses, authored by Natalia Menéndez González, Francesco Paolo Levantino, Sofia Verza and Stefano Trevisan, and published in the Computer Law & Security Review.
Natalia Menéndez González is a Research Associate at the Centre for a Digital Society of the EUI’s Robert Schuman Centre. She holds a PhD from the European University Institute, with a dissertation titled ‘Cultured Proportionality: Navigating the EU’s Fragmented Facial Recognition Landscape.’
Sofia Verza is a Research Fellow at the Centre for Media Pluralism and Media Freedom (CMPF) at the EUI’s Robert Schuman Centre. As part of the “Media Pluralism Monitor” project, she focuses on safety of journalists, surveillance, SLAPPs and the independence of media authorities in Europe.