A - I n f o s

a multi-lingual news service by, for, and about anarchists **
News in all languages
Last 30 posts (Homepage) Last two weeks' posts Our archives of old posts

The last 100 posts, according to language
Greek_ 中文 Chinese_ Castellano_ Catalan_ Deutsch_ Nederlands_ English_ Francais_ Italiano_ Polski_ Português_ Russkyi_ Suomi_ Svenska_ Türkurkish_ The.Supplement

The First Few Lines of The Last 10 posts in:
Castellano_ Deutsch_ Nederlands_ English_ Français_ Italiano_ Polski_ Português_ Russkyi_ Suomi_ Svenska_ Türkçe_
First few lines of all posts of last 24 hours

Links to indexes of first few lines of all posts of past 30 days | of 2002 | of 2003 | of 2004 | of 2005 | of 2006 | of 2007 | of 2008 | of 2009 | of 2010 | of 2011 | of 2012 | of 2013 | of 2014 | of 2015 | of 2016 | of 2017 | of 2018 | of 2019 | of 2020 | of 2021 | of 2022 | of 2023 | of 2024 | of 2025

Syndication Of A-Infos - including RDF - How to Syndicate A-Infos
Subscribe to the a-infos newsgroups

(en) France, OCL CA #351 - AI Act - Insufficient Protection (ca, de, fr, it, pt, tr)[machine translation]

Date Mon, 14 Jul 2025 09:55:05 +0300


Since February 2, 2025, some of the provisions of the EU Regulation on Artificial Intelligence (AI Act) have begun to apply. The result of a rather long (1) and complex process, this European agreement is intended to regulate the uses of artificial intelligence and protect individuals. We will see that AI poses a set of dangers from which the AI Act provides very little protection. ---- This article is part of the follow-up to the dossier on the computerization of the world, published in issue 349 of Courant Alternatif (April 2025), and in particular the first article on AI. ---- This European regulatory framework claims to pursue the following five groups of objectives:

ensure that AI systems placed on the market are safe and comply with current legislation on fundamental rights, EU values, the rule of law, and environmental sustainability; Promote the adoption of human-centric and trustworthy AI;
Ensure a uniform legal framework to facilitate investment and innovation;
Strengthen governance and effective enforcement of existing legislation on security requirements for AI systems and fundamental rights;
Improve the functioning of the internal market for legal and safe AI applications, and prevent market fragmentation.
Obviously, it's easy to see that the primary purpose of this legislation is to promote markets, investment, and security. Given that AI, by "crunching" much more data than traditional search engines, consumes increasing amounts of electricity and water for cooling systems, talking about "environmental sustainability" is a joke in poor taste.

But what concerns us even more, despite this display of respect for values, are the dangers of AI for fundamental freedoms and human rights. Amnesty International (2) has identified some positive points and, above all, some major dangers. The first positive point is the ability to file complaints and seek compensation for individuals who may have been victims of AI-based actions. The second is the ban on "social rating" software, whether public or private. For the time being, we should avoid the Chinese model (3).

The European regulation on AI has several flaws, and in particular, as a Disclose investigation (4) revealed, it was the French government that actively campaigned to weaken the text's scope.
The AI Act prohibits a whole list of uses of AI deemed "unacceptable," but it often adds details on cases allowing exceptions. Here are three examples:

AI systems intended to assess or predict the risk of a natural person committing a criminal offense are prohibited... unless it is to support a human assessment of a person's involvement in criminal activity. Basically, these AI systems for police use are prohibited if they are used predictively (having a certain profile, this person may commit an offense), but they are authorized for confirmatory purposes (this person, suspected of having committed an offense, has a certain profile, so there is a high probability that they actually committed it). This promises some tough battles in court if this AI profiling is used to compensate for the lack of evidence.

AI systems for emotion recognition are prohibited in the workplace and educational institutions... except for medical or security reasons.

Real-time remote biometric identification systems are prohibited in publicly accessible spaces for law enforcement purposes... except in cases of targeted searches for potential victims of crime, prevention of a specific threat to the life or safety of persons or prevention of a terrorist attack (5), or identification, location, or prosecution of perpetrators or suspects of certain criminal offenses punishable by a maximum sentence of at least four years.

Facial recognition is only partially prohibited. The AI Act paves the way for the legalization of certain specific uses of real-time facial recognition. This is a first in the EU, and it constitutes a violation of rights and freedoms.
AI is authorized at borders to monitor exiled persons. In particular, while the AI Act prohibits "emotion recognition" and "biometric categorization" in workplaces and schools, their use by law enforcement and immigration authorities is permitted. The AI Act therefore authorizes states to test and use dangerous surveillance technologies at EU borders.

As a result, one might assume that the EU is ready to test abusive surveillance systems on society's most marginalized populations and could then extend them to other population groups.

This regulation is applicable only within the EU; nothing is concretely planned to protect European populations against the abusive use of prohibited systems by foreign firms. In particular, the regulation classifies AI technologies into four risk levels: unacceptable risk, high risk, transparency risk, minimal or no risk. As good liberals, European countries have decided that the private sector, following intense lobbying, will define the risk levels: AI system providers will themselves define whether their tool should be considered high risk or not.

Finally, states retain complete latitude to use AI as they see fit, regardless of individual rights. As stated in Article 2.3 of the AI Act, "this Regulation (...) does not affect the competences of Member States in matters of national security." Similarly, Article 46, paragraph 2 of the European Regulation specifies that "in a duly justified emergency situation for exceptional reasons of public security (...) law enforcement authorities" may use "a specific high-risk AI service" without authorization.
It is therefore absolutely necessary to denounce the rhetoric presenting the AI Act as a regulation protecting freedoms. We must also denounce AI as a technology that, in addition to being environmentally and socially costly, constitutes a threat to individual freedoms. And to achieve this, it is necessary to combat all the techniques (video surveillance, computer tracking, files, etc.) that already fuel AI. In addition to the denunciations, it seems that there are increasing acts of sabotage of cables and other infrastructure. This can't be a bad thing. We can also practice digital hygiene individually by using spying tools as little as possible, or by using systems that are slightly more secure than the most widespread ones.

AD, Limoges

Notes
1) The basic text was adopted in December 2023, the agreement voted on in March 2024, and ratified in June. The provisions were to be gradually implemented between August 2024 and August 2026.
2) Several articles on their website, but particularly the one entitled "How to Regulate Artificial Intelligence."
3) Watch the documentary "My Wife Has Credit"
4) On Disclose.ngo, "Artificial Intelligence: France Opens the Way to Mass Surveillance in Europe"
5) As some police investigations against activists (the "8/12" and "15/06" cases) were conducted using counterterrorism resources, we can be sure that exceptions will be frequently activated.

http://oclibertaire.lautre.net/spip.php?article4457
_________________________________________
A - I N F O S N E W S S E R V I C E
By, For, and About Anarchists
Send news reports to A-infos-en mailing list
A-infos-en@ainfos.ca
Subscribe/Unsubscribe https://ainfos.ca/mailman/listinfo/a-infos-en
Archive: http://ainfos.ca/en
A-Infos Information Center