Agreement between major groups in the European Parliament on AI regulation has been lost, leaving the door open for amendments from both sides.
The AI Act is a landmark law regulating artificial intelligence based on its potential to cause harm. The European Parliament is scheduled to vote on the bill on June 14, as the deadline for submitting amendments has passed on Wednesday (7 June).
At the end of April, the four major political parties agreed not to submit any alternative amendments, with the exception of some of the European People’s Party (EPP), which was granted some flexibility on the remote biometrics issue.
But the EPP abused such flexibility against other major groups by submitting individual amendments as a group rather than in a split vote, which would have resulted in individual votes on the terms of the compromise.
“EPP has scrapped the deal. They need to take responsibility for it. This deal was not a group amendment,” said a European Parliament official, as other groups decide to back other amendments. “Anything is possible,” he added, because of the possibilities.
Center-right parliamentary officials called the accusations “fake news” and claimed that “the minutes show that the EPP was granted flexibility with respect to the RBI.” [remote biometric identification] at the plenary session. It was always up to our group to decide how to do it. ”
remote biometrics
The issue of remote biometric authentication in public spaces has been a hot topic in parliamentary debates, with lawmakers battling between the need for security and the risks of mass surveillance.
A compromise would have been to ban real-time use of the technology while allowing post-mortem investigations of serious crimes with the approval of judicial authorities.
Rep. Jeron Renners texted to coordinate EPP amendments to allow real-time use of these systems in three exceptional cases: finding missing persons, preventing terrorist attacks, and searching for serious crime suspects. essentially reverted to
The text stipulates that these are specific cases requiring prior approval by judicial or independent administrative authorities and are subject to protective measures on the basis of necessity and proportionality.
In contrast, the Left has submitted an amendment that would outright ban remote biometric technology. The proposal may find support among political groups who feel they are being wronged by the EPP.
Prohibited Actions
AI regulations prohibit applications deemed to pose an unacceptable risk to society. In response to impulses from left-to-centrist parliamentarians, this list has expanded to applications such as predictive policing and sentiment recognition.
Individual lawmakers have tabled several amendments that are part of the political agreement and would further expand the list of banned AI applications.
A bipartisan coalition including Pirates MP Patrick Breyer, Social Democrat Birgit Schippel and liberal Karen Melchior bans AI systems from detecting, monitoring and analyzing people’s behavior in public. Submitted an amendment.
“Behavioral analytics is a form of mass surveillance of public spaces that automatically alerts authorities to ‘abnormal behavior’. Such practices educate conformist behavior,” the justification reads.
Scipel also asked Congressman Sylvie-Guillaume to profile or assess whether immigration could be a threat based on known or predictive data on individuals, or to predict individual or group movements regarding border crossings. It submitted two separate amendments to ban AI-powered tools for
Left-wing lawmakers support both amendments, and have also proposed banning systems that could be used to detect the presence of people in workplaces, schools, border surveillance, and elsewhere.
high risk system
The two political groups that did not support the deal were the Left and Conservative Reform Europe, who submitted 16- and 9-point amendments, respectively.
The main amendments relate to classifying AI systems as high risk of harm, requiring developers to follow stricter obligations on risk management, data governance and technical documentation in such cases.
Textually, AI systems that fall into a list of critical areas or use cases are automatically qualified as high risk. This automation was removed as a concession to the center-right, but the left is now trying to reintroduce it.
This infringement document significantly refines and refines this list of high-risk areas and use cases, including recommendation systems used by social media platforms designated as systematically relevant under the Digital Services Act (DSA). Expanded.
Conservative amendments limit this category to platforms that are not DSA compliant, explaining that “there is no need for duplication of obligations.”
Generation AI
For underlying models, especially generative AI like ChatGPT, MEPs agree to introduce certain obligations, including the fact that they must publish sufficiently detailed overviews of their training data that are subject to copyright law. Did.
Conservative lawmakers have proposed introducing warnings about whether such summaries are subject to the exclusion of data protected by intellectual property rights or constitute trade secrets.
Additional Rights
The left-wing amendment calls on public interest groups to require those affected by AI systems to seek clarification of their decision-making processes and to lodge complaints with the competent national authorities if they believe the AI system violates AI law. A right to request is also introduced.
[Edited by Nathalie Weatherald]
