Germany · de.indymedia · · 3h
Too big to fail: Irresponsible AI panic
Deutsch (original) · Auto-translated to English
For years, the AI industry has been warning about the dangers of the technology that it has unleashed on the world. To date there has been little improvement. This is also because the tech companies determine the framework of the debate. It's time for them to take real responsibility. A comment.
They independently break into government websites, hack foreign companies, or contribute to schoolchildren being torn apart in bombing raids. Even the heads of large AI providers such as OpenAI or Anthropic are no longer entirely comfortable with their own products. Just last week they warnedhttps://netzpolitik.org/2023/offener-brief-zu-ki-opfer-des-hypes/">renew...about the dangers of technology, this time before the UN Security Council, and called for international cooperation to contain them.
We should slowly take them seriously, but not at their word. Whether the AI apocalypse is really imminent as some of them claim,https://netzpolitik.org/2026/warning-vor-katastrophalen-schaeden-kuenstl...left open. But the biggest possible accident doesn't have to happen in order to clearly understand the destructive potential of AI - and especially of AI companies themselves.
Without regard for personal data, copyright or the people who will suddenly live next to data centers, these companies have absorbed virtually all of humanity's digitized knowledge. They package this into commercial business models andhttps://finance.yahoo.com/news/sam-altman-says-ai-eventually-152157902.h...us this “intelligence”back again. It may be the “largest theft of labor in human history,” a Microsoft executive wrote in an internal memo filed in a lawsuithttps://www.washingtonpost.com/business/2026/09/17/microsoft-exec-called...Light came.
In just a few years, these companies have become deeply embedded in our societies with breathtaking speedhttps://www.ft.com/content/00f94018-e658-4545-b16e-1bc00e19b754">writecarry on cheerfully. Often enough they spur on leading politicians, from US President Trump to the former British Prime Ministerhttps://netzpolitik.org/2025/britische-regierung-ki-in-die-venen-der-nat...up to the President of the EU Commissionhttps://netzpolitik.org/2026/rede-zur-lage-der-union-wir-wollen-kuenstli...the Leyen. You don't want to miss the connection, even if you find yourself therehttps://netzpolitik.org/2025/kuenstliche-intelligenz-ursula-von-der-leye...the tech bossesmight.
A bet without us
It's a gigantic bet on future markets, business models and revenues that these companies are making. It is anything but certain that theirshttps://www.ecb.europa.eu/press/blog/date/2026/html/ecb.blog20260817~754...wear business models at all. One estimate estimates that the planned expenditure for the construction of data centers in the USA alone would correspond to an average of 3.6 percent of the US gross domestic producthttps://www.brookings.edu/articles/financing-the-ai-buildout/">currentstudy. Never before has the US economy been so dependent on the expansion of a single industry, warnshttps://www.wsj.com/economy/the-ai-build-out-is-becoming-the-biggest-eco...Street Journal.
Former heads of the US Federal Reserve are now movinghttps://www.washingtonpost.com/opinions/2026/09/03/ai-safety-security-ri...to the last financial crisis. At that time, inadequately regulated banks brought the international financial system to the edge of the abyss with high-risk business models. They, who had become too big and too important, were ultimately saved with tax money. The rest of society has to bear the costs.
Too big to fail: This seems to be one of the mottos with which today's tech oligarchs use theirshttps://www.nytimes.com/2026/09/26/opinion/ai-bubble-banking-crisis-too-...secure the endwant. However, they do not want to take responsibility. For example, Palantir boss Alex Karp recently suggested that, in his view, corporatizing AI companies would be the right approach.
–––
The hut is on fire! We're going in.
We oppose fascisation. We fight for a free, open and solidarity-based society. We defend basic rights for everyone.https://netzpolitik.org/spenden/?via=rss">Thisonly possible with your support.
–––
In view of the increasing scare reports about AI, there would be nothing else to do, because “if you don’t nationalize them, every single one of my customers will sue.”https://www.cnbc.com/2026/09/17/ai-safety-palantir-karp.html">saidCarp. This risk is apparently acceptable to democratic states, but not to private companies. It goes without saying that they want to keep their profits for themselves.
Watered down EU rules
In her State of the European Union speechhttps://netzpolitik.org/2026/rede-zur-lage-der-union-wir-wollen-kuenstli...the CDU politician Ursula von der Leyenrecently pointed out the dangers of self-learning AI models, and in the same breath praised the AI Act as an “important step towards a safety net”. What she left under the table: It was only in May that the EU watered down and calculated the AI lawhttps://netzpolitik.org/2026/ki-verregulation-eu-parlament-und-rat-einigen-...limited for high-risk AI applications. The President of the Commission also did not mention that the Commission submitted its own proposal last yearhttps://www.europarl.europa.eu/legislative-train/theme-a-europe-fit-for-...withdrawnhas.
The bill had been on the table since 2022, but ultimately could not be opposed by two or sohttps://verfassungsblog.de/anatomy-of-a-fall-aiact-aild-pld/">at the same timenegotiated proposalsclaim: the AI Act and an updated, more general product liability policy. Actually, the three laws should work together and complement each other.
The AI Act contains rules for AI manufacturers and providers that vary in severity depending on the risk of the planned AI product, but do not provide injured users with any legal recourse. The company's own directive was supposed to cover AI-specific liability issues, but it was later absorbed into the general EU directive. Although this has been expanded to include damage caused by software, among other things, it still contains gaps - and what's more, it still awaits implementation across the EU, for example in Germany.
The fact that this framework is not sufficient to catch AI tools gone wild or, better yet, to ensure that they do not cause real harm in the first place is due not least to theIT security incident involving OpenAI and Hugging Facedisclosed in the summer: At that time, AI agents developed by OpenAI broke out of an insufficiently isolated environment during a test phase and independently attacked the infrastructure of the company Hugging Face. Since this was not a product available on the market, it is still controversial even among experts as to whether the EU rules would have covered such incidents at all. If in doubt, probably not.
Liability risk creates caution
Two responses to this problem currently appear to be emerging at EU level. On the one hand, there are proposals for international cooperation that could be institutionalized along the lines of the International Atomic Energy Agency. OpenAI boss Altman had called for this years agohttps://www.politico.eu/article/karsten-wildberger-ford-globale-ki-be...Wildberger(CDU) and probably also his party colleague von der Leyen would be on board.
–––
The work of netzpolitik.org is financed by donations from our readers.
Become part of this unique community and support our public interest-oriented, advertising- and tracking-free journalismhttps://netzpolitik.org/spenden/?via=rss">nowwith a donation.
–––
On the other hand, things are seething in the EU Parliament. A group of MPs recently formed therehttps://www.politico.eu/article/eu-lawmakers-ai-apocalypse-warnings-brus...revivewants. Parliament cannot propose laws, but it can build up political pressure. Whether this will be enough to correct the EU Commission's course of deregulation remains unclear for the time being.
Neither approach is necessarily hopeful: it could take years until a compromise is found that everyone is happy with. There is no guarantee of success. Resistance to excessively strict requirements is likely to come from both the USA and China, who are unlikely to be persuaded for strategic reasons.
In the worst case scenario, our society is threatened with being run over twice: by technology that was rolled out prematurely and has no place in the mass market without guardrails. And from a business-friendly policy that had the ability to intervene in a timely manner. If she hadn't lost her courage at the last moment.
Regulation under democratic conditions
It's not too late yet. As the AI bosses themselves demand, they and their companies must finally take responsibility - but not on their terms. They must be held liable for all the damage their products cause, without any singularity. Anyone who expects to be sued will think twice about releasing a half-baked and potentially dangerous product into the world.
To achieve this, the EU must introduce stricter regulations that can achieve this. The current situation, which is overflowing with bad news and warnings about bad news, while parts of politics are still naivehttps://netzpolitik.org/2023/gegen-den-ki-hype-die-grosse-science-fictio...worthy promise of innovationrunning after it must come to an end. Clear guidelines and real rules are needed so that billion-dollar corporations are held liable - and society is not left in ruins later on. Because that is irresponsible.
–––
The work of netzpolitik.org is financed by donations from our readers.
Become part of this unique community and support our public interest-oriented, advertising- and tracking-free journalismhttps://netzpolitik.org/spenden/?via=rss">nowwith a donation.
Read the full story at the source
Source: de.indymedia