A missed opportunity: European Parliament upholds facial recognition ban, but fails to properly protect people’s rights in AI Act vote.

14 June 2023 | Artificial Intelligence, Statement

In today’s Artificial Intelligence Act vote, MEPs failed to fully protect society at large, not to mention vulnerable groups like migrants and refugees, and voted against measures to introduce proper complaints and representation systems for human rights violations.

The long-negotiated AI Act is intended to be the world’s first major legislation to regulate against the harms Artificial Intelligence brings. The European Civic Forum has supported the European Institutions’ initial stated aim to foster trustworthy and human-centred AI that fully respects fundamental rights. 

However, we are frustrated that Parliament’s vote still falls short of this aim. While MEPs voted to approve several fundamental rights protections, including a comprehensive ban on real-time remote biometric identification, the adopted text still lacks vital safeguards. 

The version of the AI Act adopted today is clearly insufficient to address the negative impacts of Artificial Intelligence systems that we are already seeing, let alone those that will come in the future.  

“Notwithstanding the positive elements it includes, the outcome of today’s vote sadly shows that the European Parliament was unable to uphold the highest international human rights standards. Fundamental rights violations could go unchecked and it does not prohibit systems used to target vulnerable groups such as migrants and refugees,” ECF vice president Jan Robert Suesser said.

The adopted text still lacks prohibitions on harmful AI systems used in migration context which violates the rights of migrants and refugees at the borders of the EU. But the issues are not limited to this. 

The adopted text does not include explanation and complaints mechanisms that would safeguard peoples’ rights and  society at large. Individuals that want to raise concerns and complaints about AI systems that have been harmful to them are not able to seek support from a civil society organisation and mandate it to represent them and make a complaint to the supervisory authority. Neither does the proposal give rights to civil society to raise general concerns to prevent violations of people’s rights.

The co-rapporteur for the AI Act, Dragoş Tudorache stated in the Parliament’s press conference after the vote that the Parliament aimed to have a “bottom-up” approach when drafting the legislation. To make this “bottom-up” approach a reality, much has still to be done so that the voices of the civil society actors who represent people’s concerns are fully taken into consideration in the final text.

“Today MEPs voted for an act which does not properly protect society at large. To live up to its goals, the legislation would need to include efficient mechanisms for complaints and public interest representations for people affected by AI systems. Civil society actors need to be fully involved in processes that will protect people from harmful AI,” ECF’s policy and capacity building officer Kerttu Willamo said.

The result of today’s vote is that a promising law to guard against AI’s violations of human rights still fails to meet its goal. If we miss this historic opportunity, it would have serious real-world consequences. 

The EU will be a world leader on human-centred AI regulation if it takes bolder measures. The European institutions must think again and strive for a comprehensive and rights-respecting regulatory framework for AI. Civil society actors need to be given a real role in guarding against the harms AI can pose to individuals and society at large. These are benchmarks to be included in the upcoming trilogue negotiations.