Diskurs

Matthias Hornschuh:

In early 2024, the German NGO Lobbycontrol drew attention to the worrying proximity of big tech companies, and in particular a supposed German AI unicorn, to the highest levels of politics and administration in Europe and Germany[1].

In September 2024, the German government provided information on the type and number of meetings between the federal government and Microsoft and SAP in its response to a brief parliamentary enquiry from the opposition. It became clear that there had been over 100 meetings between the heads of the government and the corporations within two and a half years. A dozen of these were in the personal presence of the Chancellor, while Minister for Economy Habeck was present barely less. This led to a murmur in political Berlin. In view of the direct lobbying access of corporations to the government, there was concern about Germany's digital sovereignty to be heard [2].

At the beginning of January 2025, Corporate Europe Observatory (CEO), a Brussels-based NGO, eventually came forward. To describe their analysis of the participation processes for the practical implementation of the AI Act, which are currently unfolding in Brussels, as worrying would be to trivialize them:

‘Behind closed doors, corporations are writing rules that will govern their own AI products – including on how to prevent fundamental rights violations of European citizens. […] Independent experts and civil society are out-numbered, under-funded, and struggling in the face of the corporate dominance"[3]

The lengthy article neatly dissects the impenetrable lack of transparency of the process and the cast, the structural entanglements at national, European and international level, the dramatically different resources of corporate and civil society representatives in the absence of state support and, last but not least, the all-dominant diffusion strategies of the corporations. According to the quintessence, even before the work on the Code of Practice was completed, these had led to the societal implications of the use of artificial intelligence being categorized as technical problems. In this way, the technical process is made the object instead of considering its result; ultimately, the companies can decide on the rules for their own behaviour - including encroachments on fundamental rights. Justified by the needs and profits of the corporations.

This raises questions. But the answers remain unclear. One thing is clear: the rights of citizens, who are always more than ‘just’ consumers, are not the focus here. Nor is their enforceability:

‘The expected final result will be a checklist that corporations can use to demonstrate their compliance with the AI Act. But the “basic rules” will not be binding"[4]

Three wake-up calls. Not the only ones. Unfortunately, none of the calls woke up anyone. Fundamental rights, it seems, are currently rather fundamentally up for discussion, while few keywords in Germany lead to a similarly disinterested shrug of the shoulders as ‘digital sovereignty’ of all things. In a mixture of lethargy, rigidity of tolerance and a pronounced inferiority complex, most recently often described as FOMO, Fear of Missing Out, people move somewhere along the continuum between ‘will go away’ and ‘don't do anything’ ... and then consequently do: nothing.

But precisely where society is signaling disillusioned disinterest, there is plenty of room for aggressive lobbying. And where no resistance is to be expected, lobbying is all the more uninhibited.

In the US, the big tech powers no longer need lobbying in principle, having simply bought their way into government.[5] Back in September, Vice President-elect JD Vance blatantly threatened the Europeans - in the event of further regulation of Xitter.

The extent to which this was understood as a wake-up call remains unclear for the time being. In any case, there is still no sign of a clear response from the EU in terms of digital regulation. Not to mention a consistent implementation and enforcement of applicable law and agreed rules, be it towards the USA and China, or internally, in the work of the AI Office and the process and compliance deficiencies described above.

Participation without influence

The Authors’ Rights Initiative (Initiative Urheberrecht / IU) is part of the expert group on the Code of Practice (CoP), which is negotiating the (currently voluntary) rules in the area of copyright and transparency obligations under Chatham House Rules, among other things. This process is strongly characterized by non-European participants, there are inexplicable special rights for AI providers and a time pressure that can hardly be managed without a legal department of several people. Despite the efforts of copyright organizations and associations of other rightsholders to communicate their positions in detail orally and in writing, the current second draft does not show that the interests of rightsholders have been taken into account. Moreover, the recently published first draft of the AI template in this form is also useless for rightsholders.The Copyright Initiative is part of the expert group on the Code of Practice (CoP), which is negotiating the (currently voluntary) rules in the area of copyright and transparency obligations under Chatham House Rules, among other things. This process is strongly characterised by non-European participants, there are inexplicable special rights for AI providers and a time pressure that can hardly be managed without a legal department of several people. Despite the efforts of copyright organisations and associations of other rights holders to communicate their positions in detail orally and in writing, the current second draft does not show that the interests of rights holders have been taken into account. Moreover, the recently published first draft of the AI template in this form is also useless for rights holders.

Effective law enforcement, which the copyright regulations should enable according to the will of the European legislator, is in any case not possible under the currently planned regulations (the Code of Practice and the transparency requirements on copyright that are binding for all AI providers).

The reactions of the associations unanimously confirm this; the European umbrella organizations of the cultural, creative and media industries have also recently spoken out loud and clear in this regard: ‘If the draft is not substantially improved, the Code will fail to meet the objectives of the AI Act and directly conflict with EU law, including copyright law. Such an outcome would be unacceptable, and we cannot support a process that undermines the very principles it was designed to uphold.’ [6]

Digital sovereignty requires sovereign thinking and action

European thinking about digitality is strongly characterized by a focus on technology; in our dominant conception of innovation, it is virtually equated with technology. However, regulatory considerations of technology are meaningless if the role of its developers and providers is not taken into account. And considerations of digital cultural techniques (and that is precisely what we are talking about) are dubious if they do not take society and its culture into account - and thus the technological consequences that need to be assessed.

In view of a society that tends to identify innovation without justification or intention, and in which growth is an almost unconditional priority, it is hardly surprising that the EU is clearly prepared to take paths that prioritize technologically determined progress and economic growth as goals to be achieved when establishing political procedures for technology control. The EU's internal market mandate does the rest. However, this does not justify the lax approach to fundamental rights, civil rights and copyright.

There is nothing wrong with technology; this text is not carved in stone either. And nothing against progress ... But what exactly is progress? Innovation? Are we still on the same page on that? At Re:publica 2024, computer scientist and philosopher Jürgen Geuter relentlessly dissected the largely meaningless concept of innovation[7]. Generative AI in particular is extremely unsuitable for upholding it: No, this technology in particular will not cure cancer or save the climate. Incidentally, given the immense hunger for resources of generative AI, the second promise in particular is particularly bizarre evidence of the essentially irrational, downright religious sales strategy of the corporations, for whom the current backlash for supposedly left-wing green-woke misgivings fits in perfectly.

Nevertheless, it is the narrative of linear innovation, which has been repeated incessantly for three decades and which unfolds in an evolutionary and therefore unstoppable manner, behind which the doubters (aka ‘deniers of progress’, ‘luddites etc.) supposedly threaten to fall behind, to which German and European policy makers expose themselves. Over and over again; even the privileged liability of service providers in the e-commerce directive at the end of the 1990s was due to concerns about falling behind. And time and again, it was the authors, performing artists and their exploiters and producers who had to pay the high price for a policy that degraded existing law to a nice-to-have.

It would therefore certainly be a step forward for the EU not to give in to the FOMO that is constantly being drummed into it. Fear is a bad adviser - and undoubtedly an enemy of sovereign action.

Sovereignty at state level describes the guarantee of uninfluenced and independent freedom of decision and action, whereby a distinction must be made between legal and political science definitions. Politically, sovereignty at the beginning of the 21st century ‘is understood in many states as the sovereignty of the people: Sovereignty emanates from the people alone.’ In a representative democracy such as the German one, ‘the operative political events are in the hands of a parliament.’ In the legal sense, the sovereign state ‘has the sovereign power to enact acts that are binding for all those subject to its territorial or personal authority [...]. It has the right of final decision in all public matters [...]"[8].

It would be interesting to know to what extent the EU and its member states are actually capable of acting independently of non-European technology providers. In autumn 2024, the EU Parliament announced that it’s decision to rely on Anthropics AI Claude[9] for the exploitation of existing databases in the future. The German government is known to pay around €200 million a year in license fees to Microsoft alone, which gives an idea of the extent to which Microsoft products and services are used by the federal government ... for example, to write and negotiate draft laws to control and restrict the market power of big tech companies.

Those to be regulated are therefore the providers of the technologies that are used to develop and negotiate the regulation. It remains to be seen to what extent the AI model built into Microsoft's Office package reads and calls home as standard, but a significant step towards sovereignty would be the decisive use of European services - and, if necessary, their commissioning or development. There are also repeated calls for open source software solutions (which, incidentally, have their own problems from a copyright perspective, but that's another topic).

To come back to the question of sovereignty: Europe must guarantee and defend its sovereignty in shaping the digital. This independence should not be based on technology, but should be conceived by people and managed by them; it should have people at the centre of all intentions and at the lever of all instruments. This would be the antithesis to the rampant techno-solutionism that promises technical solutions to (supposedly) technical problems and, in an attempt to contain corporate overreach, cedes power to the corporations. The factual and practical independence of our decisions will therefore prove to be independence from the determination of the neo-feudalistic masters of technology, the tech bros.

It is unacceptable that these same feudal lords not only sit at the table when their regulation is being negotiated, but even dictate the menu. Digital sovereignty means being able to draw up your own rules on your own terms. Not letting those to be regulated write the rules that should apply to them.

Transparency: achievable!

Now you migtht object: But the people in politics and administration have no idea about the technology; feasibility must be guaranteed! Good point: this technology supposedly cures cancer and climate change. It will be much more intelligent than humans this year, or next year, or any other year, we'll see. Which, when the time comes, will be discovered by having it solve multiple choice A-level questions (which were included dozens of times in the training corpus, along with the solutions; but nobody out there needs to know that).

It is such a fantastic, almost magical technology that - and this is no irony - is such a powerful master of memorization and retrieval that some computer scientists describe it as a knowledge database, contrary to pure doctrine. She can surely be persuaded to help her developers reveal what she has learnt. And if not, then their masters could simply let them develop a solution for their own problem. I hear they can code…

Sorry, my enthusiasm just got the better of me. But this is all a serious topic. Quite unironically. Because it's about the future possibility of paid (erwerblich) creative activity in culture and the media at a professional level. Our colleagues from the German Federation of Journalists (DJV) wrote the following on 31 January 2025:

‘The German Federation of Journalists is calling on the EU Commission's AI Office not to allow itself to be influenced by the big tech industry and to protect the legitimate interests of authors. The background to this is the draft template for implementing the transparency obligations from the AI Act for AI model operators. The draft does not provide for a work-specific obligation to provide evidence. ‘The current draft of the AI template does not provide any added value for authors,’ criticizes DJV legal advisor Hanna Möllers. ‘None of the information provided will help authors in court. This contradicts the meaning and purpose of the AI Regulation, which expressly states that the disclosure obligation should make it easier for rights holders to enforce their rights.’

The big tech industry, on the other hand, is claiming trade secrets. ‘The publication of plant-specific information does not violate trade secrets,’ explains the DJV lawyer. ‘Nor does it harm food manufacturers if they have to disclose ingredient lists. Only the recipe, not the ingredients, can be a trade secret,‘ says Möllers.’[10]

Let us be clear about this: The industry that has been accused worldwide of the biggest theft in human history - specifically the theft of intellectual property - is invoking its ‘trade secret’ (aka intellectual property) to avoid having to disclose the provenance of the training data. And the few authorized representatives of the victims of theft have to write open letters of protest to save what can still be saved. There's a word for that: bigotry.

Meanwhile, the added value generated by the models trained with unpaid European content takes place almost entirely in the USA and Asia: Europe is giving away its global knowledge just to buy it back afterwards. This is not only downright idiotic, but also very lacking in sovereignty. It is an expression of colonialist submission.

Edit 05.02.2025: After this text was completed, a piece of news made the rounds that, had it arrived earlier, would definitely have been included in the text.

At a sales event in Brussels on the evening of 4 February, which entertainingly bore the perfectly fitting and incredibly original name #EUInnovationDay, Meta ‘Policy Lead’ Joel Kaplan expressed his incomprehension that the EU was still clinging to its cranky idea of AI regulation. Meta is apparently not going to sign the Code of Practice.

Which, if it turns out to be true, would undoubtedly be the first good news from the USA after the gruelling first 15 days of Trump2.

[1] https://www.lobbycontrol.de/macht-der-digitalkonzerne/ai-act-neue-dokumente-zeigen-grosse-nahe-zwischen-aleph-alpha-und-der-bundesregierung-113607/ Accessed on 02/02/2025

[2] https://mdb.anke.domscheit-berg.de/2024/09/microsoft_lobby/ Accessed on 31/01/2025

[3] https://corporateeurope.org/en/2025/01/bias-baked Accessed on 02/02/2025

[4] Ibid.

[5] https://urheber.info/diskurs/auf-konnvergente-topfe-gehoren-konvergente-deckel Accessed on 03/02/2025

[6] https://www.enpa.eu/joint-letters/right-holders-joint-letter-on-the-second-draft-code-of-practice Accessed on 04/02/2025

[7] https://tante.cc/2024/05/28/empty-innovation-at-republica-2024-rp24/ Accessed on 02/02/2025

[8] https://www.staatslexikon-online.de/Lexikon/Souveränität Accessed on 05/02/2025

[9] https://www.linkedin.com/posts/daniela-amodei-790bb22a_the-european-parliament-is-using-anthropics-activity-7254172155595169793-tNUc/ Accessed on 03.02.2025

[10] https://www.djv.de/news/pressemitteilungen/press-detail/djv-fordert-ausgewogene-transparenzpflichten/ Accessed on 03/02/2025