grandeduc - stock.adobe.com
The European Parliament has voted in favour of a proposal to allow international crime agency Europol to more easily exchange information with private companies and develop artificial intelligence (AI)-powered policing tools.
The proposal was put forward by the Parliament’s Committee on Civil Liberties, Justice and Home Affairs (LIBE), and makes notable extensions to Europol’s mandate in three main areas, including its cooperation with private parties, its processing of personal data in support of criminal investigations, and its role in research and innovation.
However, civil rights groups say the proposed mandate represents a “blank cheque” for the police to create AI systems that risk undermining fundamental human rights.
Under the proposed mandate – which was passed with 538 votes to 151 – Europol will have the capacity to process data provided by any private company, including big tech firms, as well as any third country that voluntary hands the data over.
Europol will also be able to identify themes for EU research programmes, formalising its ability to train algorithms and develop new tools for use by law enforcement throughout the bloc, as well as act as a processor for large quantities of data itself.
A separate proposal amending the Schengen Information System (SIS) regulation to enable Europol to make third country data available to frontline officers was also passed by the Parliament, with 545 votes to 147.
By giving the Europol the ability to submit its own SIS alerts, European national authorities will no longer need judicial authorisation to access the agency’s data, as it will be empowered to make its own entries into the EU-wide database.
Following the vote, the European Commission (EC), Parliament and Council must now negotiate a final version of Europol’s mandate between them.
The European Council agreed its negotiating mandate in June 2021, which is broadly in line with the proposed changes just voted on by the Parliament.
MEPs contradict previous vote opposing certain uses of AI by police
MEPs in the LIBE committee previously voted in favour of the proposal to extend Europol’s mandate on 12 October 2021, with 47 for and 16 against. Out of those who backed the proposal, all belonged to the European People’s Party (EPP), centrist Renew or centre-left Socialists & Democrats delegations, while those who voted against were from the Green or Left delegations.
However, a week earlier on 5 October, the Parliament approved a LIBE committee report on the use of AI by police in Europe, which opposed using the technology to “predict” criminal behaviour and called for a ban on biometric mass surveillance.
While some centre-right members of LIBE from the EPP group pushed for three amendments to the committee’s report that would have made it easier for European police to potentially conduct predictive analytics and biometric surveillance – which civil society critics claimed at the time would have weakened the bloc’s commitment to fundamental human rights if accepted – they were also voted down by the wider Parliament on 5 October.
The full text of the LIBE report clearly identifies that many AI-driven identification techniques in use today “disproportionately misidentify and misclassify, and therefore cause harm to racialised people, individuals belonging to certain ethnic communities, LGBTI people, children and the elderly, as well as women”.
It further “highlights the power asymmetry between those who employ AI technologies and those who are subject to them”, and calls for outright bans on the use of AI-related technologies for proposing judicial decisions; for any biometric processing that leads of mass surveillance in public spaces; and for the mass-scale scoring of individuals.
Addressing the European Parliament on 21 October 2021 ahead of the mandate vote, Green MEP Saskia Bricmont said: “Authorising Europol to develop research and innovation artificial intelligence projects for police and experimental uses, and promoting technology for facial-recognition, is something that the Parliament opposed just two weeks ago.”
She added that the “commission is putting forward a reform that would legalise the illegal practices” of Europol, referring to the European Data Protection Supervisor’s (EDPS) September 2020 decision to admonish the agency for its data processing practices.
Specifically, the EDPS found Europol went beyond its existing mandate by collecting the personal data of individuals who had not engaged in any criminal activity.
A blank cheque to develop high-risk AI systems
Although Europol’s management board will now be required to nominate a Fundamental Rights Officer following the Parliament’s vote, and the EDPS will be given oversight of the agency’s personal data processing operations, Fair Trials – a non-governmental organisation (NGO) campaigning for a fairer justice system globally – has said that the new mandate will effectively give Europol a “blank cheque” to use and further develop high risk AI systems for policing.
“We are deeply disappointed and concerned by today’s vote to extend Europol’s mandate. MEPs have already recognised the serious threat that artificial intelligence in policing poses to fundamental rights. We cannot understand why they have now given Europol free rein to develop and use these same tools that reinforce discrimination and undermine human rights,” said Fair Trials legal and policy director Laure Baudrihaye-Gérard.
“All law enforcement agencies must operate with accountability and meaningful oversight. There is no reason that Europol should be exempt from these obligations.”
She added that while the EC’s proposed Artificial Intelligence Act (AIA) is still to be debated in the European Parliament, it already contains specific exemptions for Europol, meaning the agency and any AI system it deploys will not be subject to the safeguards.
These carve-outs can be found in article two of the AIA, and recital 11 of the pre-amble, which specifically mentions Europol.
Digital civil rights experts and organisations – including Access Now and European Digital Rights (EDRi) – have previously told Computer Weekly the AIA would fall short on protecting fundamental rights, claiming it would essentially act as a green light for a number of high-risk use cases due to its emphasis on technical standards and mitigating risk over human rights.
According to Baudrihaye-Gérard, the extension of Europol’s mandate in combination with the AIA’s exemptions would effectively allow the crime agency to operate with little accountability and oversight when it comes to developing and using AI for policing.
“This is a massive carve-out which would enable Europol to operate without further regulation than what is provided in its mandate, which will now be extended even further,” she said.
“Considering today’s vote, we are going down a path in which Europol is allowed to operate with little accountability or oversight. No-one is asking questions. No-one is holding the agency to account. It is deeply worrying for fundamental rights, including the right to a fair trial and the presumption of innocence.”
While critics claim the recast Europol mandate contradicts the previous vote by MEPs on use of AI by European police, its rapporteur, EPP member Javier Zarzalejos, argued this was not the case.
“The reform of Europol’s mandate is a long-standing request by this Parliament, and is mainly motivated by the need to equip the agency with new tools and capabilities to face the digital transformation,” he told MEPs ahead of the vote.
“Criminals have adapted their modus operandi to the new digital reality, and consequently it is necessary to enable Europol to better support member states in the fight against these evolving security threats.”
He added that the new mandate would introduce additional safeguards to ensure that fundamental rights are respected.
Read more about artificial intelligence
- London has published the latest iteration of its Emerging technology charter, a set of practical and ethical guidelines that outline the city’s expectations for how new data-enabled technologies should be developed and deployed for use in the public realm.
- Any attempt to regulate artificial intelligence must not rely solely on technical measures to mitigate potential harms, and should instead move to address the fundamental power imbalances between those who develop or deploy the technology and those who are subject to it, says a report commissioned by European Digital Rights.
- Algorithmic accountability policies should prioritise meaningful public participation as a core policy goal, so that any deployment actually meets the needs of affected people and communities, according to a global study of algorithms in the public sector.