Hello!
Things are becoming quieter on our side. The Artificial Intelligence file is done. We have been to Strasbourg with Ivo Kruusamägi from Wikimedia Eesti to talk to government reps and potential allies about SLAPPs.
As timing goes, the last parliamentary plenary will take place on April 22-25. This is also the last chance to pass legislation during this mandate. Everything unfinished might be continued by the next parliament, but this is not a given.
Dimi & Michele
=== AI Act ===
Member States needеа to vote on a “take-it-as-is-or-leave-it” proposal by the Belgian Council Presidency. Germany, France and Italy criticised the proposal over “over-regulation” of general purpose AI models (a.k.a. large language models). But finally, a sufficient number of Member States agreed to the deal and the three big ones decided to fall in line.
—
What's in it: The one important thing to understand is that the AI doesn’t simply regulate the technology, but often regulates specific use cases. In other words, an AI system that adds funny backgrounds to your videos won’t have the same obligations as systems that handle admissions into educational institutions, help with recruitment processes or are used as part of essential public services. The latter fall in a “high-risk” category. Their deployers will be liable to ensure human oversight, data quality, keep documentation and assist potential injured parties and authorities. They will also have to report known issues. None of the AI systems used by or on Wikimedia projects appear to be “high-risk” under the regulation’s definitions.
—
The AI Act also means to prohibit certain uses. The bans include systems aimed at influencing behaviour (manipulative or deceptive AI-aided techniques) or the use of biometric information to figure out a person’s race, sexual orientation, beliefs or trade union membership. Social scoring is also explicitly prohibited. The law in theory also doesn’t permit the use of real-time facial recognition, although there are significant law enforcement exceptions to that.
—
Another category in the AI Act is “general purpose AI” (a.k.a. Large Language Models). Think ChatGPT or Google’s Bard. Developers of such systems will be obliged to keep detailed documentation and assist anyone deploying their models in understanding the systems’ functionality and limits. Furthermore they will need to provide a summary of the copyrighted material used to train the models, which in practice includes freely licensed content. Don’t ask how this got in there, but be happy instead that that’s the only place copyright gets dropped in the file :P
—
The relevant committees approved the text. Next up is a plenary vote that is expected in April.
=== anti-SLAPP ===
On the 5 and 6 january Wikimedia Europe and Wikimedia Eesti (WMEE) took part in the in-person CASE (Coalition against SLAPPs in Europe) meeting in Strasbourg. On the first day, we attended an event at the Council of Europe hosted by the UK delegation (the CoE is working to adopt a Recommendation). Ivo from WMEE briefly presented his case and made delegates aware that also Wikipedians can be SLAPPed.
—
On the second day, the focus was on the future works of the Coalition: Member States will indeed have two years time to transpose the EU Anti-SLAPP Directive once it enters into force. This was the occasion to establish contacts, make other partners aware of the circumstance that we want to engage more on the issue of SLAPPs and give our inputs. In particular, we established contact with Matthew Caurana Galizia (the son of Daphne Caruana Galizia), who is one of the founders, and main driving force, of the Coalition.
—
As for the Directive, it was adopted on Tuesday 27 February by the European Parliament Now the text needs to be voted in Council and then published in the Official Journal before entering into force.
=== DSA & Disinfo Guidelines ===
The European Commission wants to publish guidelines by the end of March to instruct Very Large Online Platforms (so including Wikipedia) on how to deal with integrity of electoral processes, including next European Parliament Elections that will take place in June.
—
There is a call for feedback which the Wikimedia Foundation will participate in.
=== CSAM ===
The Child Sexual Abuse Regulation file will not pass. Council still can’t agree on a position. To save the situation, the European Commission is suggesting to make things like deepfakes of kids or livestreaming of child sexual abuse and the sharing of pedophile handbooks online a crime across the bloc.
—
All this comes as a proposed update to the 2011 Directive tackling CSAM (which was supposed to be replaced by the regulation. As time has run out, it will be the next European Parliament, together with theCouncil, that will be working on this.
=== Cyber Violence Law ===
EU legislators reached a provisional deal (vulgo: trilogue agreement) on rules to combat gender-based violence and protect its victims, especially women and victims of domestic violence.
—
Broadly speaking, the Directive wants to tackle things such as nude deepfakes, revenge porn, cyberstalking and online harassment . It will force EU countries to be more stringent in punishing abuse against women, including through an obligation to introduce new criminal offences (if they are not on the national books yet) and by setting minimum standards.
—
Don’t hold your breath, though! Member states have until 2027 to introduce the changes needed in national law.
=== Polish Copyright ===
Poland is finally, finally moving ahead with implementing the EU’s Copyright in the Digital Single Market Directive (remember that?!). Poland is the only country that hasn’t transposed the new rules yet. There has been a public consultation, which our Polish community (Maciej and Szymon) participated in. Here’s a rundown of their contribution in Polish and the machine translated version in English.
=== EU Elections ===
Ahead of EU elections European party groups have usually publish documents called manifestos that outline their main proposal. Last month we covered the Social Democrats and the Greens. This time we have ALDE (the pan-european liberal party that shares a parliamentary group with Macron’s MEPs). Their manifesto is technically still at a drafting stage, but it is almost finalised.
—
No surprisingly, ALDE tackles digital transitions by making “investments and innovation” a priority. The party also wants to help services scale up by removing existing barriers.
—
On the topic that seemingly no one can avoid these days - artificial intelligence - the liberals reiterate that they want to fully use the potential of this new technology but are keen on “human-centric innovation”.
—
Another sexy-sounding approach is the proposed “digital first principle” for legislation and rules. What this should mean in practice, however, remains somewhat unclear.
=== Transparency and targeting of political advertising ===
On Tuesday 27 February Parliament adopted the final text of the regulation, which now needs just to be voted by Council before being published. The new rules will enter into application after 18 months from the adoption of the regulation. This practically means that they will not be in place for the future EU elections foreseen in June 2024.
===END===