A dispute between Anthropic and the US government is causing unrest in the German startup scene. Critics see this as a wake-up call.
It’s a back and forth with far-reaching consequences: First, Anthropic declared that the new AI model Mythos could not be published. It is too dangerous and can independently detect and exploit security gaps. Then selected users were given access to a variant called Fable. Shortly afterwards this access was blocked again.
The background is apparently pressure from Washington. The US government wanted to make the technology available only to American users. Anthropic responded with a drastic decision: Then no one would get it.
The incident shows how dependent Europe has become on a few American AI companies. If access to the most powerful models can suddenly be politically restricted, this will not only affect researchers and developers, but also the business models of numerous startups.
There must be no “appeasement policy” that the Trump administration allows.
Christoph Stresing
Managing Director of Startup Association
When asked by Gründerszene, Christoph Stresing, managing director of the startup association, said Germany and Europe must now work with all their might to develop their own models. “The US government’s instruction to Anthropic to block the latest software development from foreigners is a serious turning point.” This action makes it clear that sovereignty is not a “nice to have,” but a question of “economic viability,” says Stresing.
Europe in dependence
Germany and Europe must now “do everything possible” to develop their own models. This also includes “building up your own independent AI infrastructure as quickly as possible”. “There must be no “appeasement policy” that the Trump administration allows,” demands Stresing.
One of the first reactions from the German startup scene came from Judith Dada, partner at the Visionaries Club. The incident is a “deeply disturbing precedent,” she writes on Linkedin. Whatever the US government’s purpose with its decision, Europe’s technological dependence on the USA has become obvious. What is needed is an AI strategy that is “orders of magnitude” bolder at the European level.
Dada published the Europe 2031 manifesto four days ago with co-authors such as Lily Stelling and Philip Fox. In it, the authors call for, among other things, massive investments in artificial intelligence, an AI alliance of medium-sized countries and reforms of the labor market based on the Danish model.
Digital sovereignty
The digital association Bitkom is also concerned. The lockdown has “a direct impact on the performance of our traditional industries and our administrations and also affects our security and, for example, the excellence of our science.” Digital sovereignty and your own AI skills are therefore of the highest priority, explains Bitkom President Ralf Wintergerst.
The public criticism of Anthropic itself is particularly noteworthy. The company is by no means considered an opponent of state regulation. On the contrary: Anthropic has been calling for stricter controls for powerful AI systems for years and regularly advocates for government supervision.
The US government’s concern relates to risks that would fundamentally also exist with other models, such as its competitor ChatGPT. If states want to restrict access to powerful AI, this must be justified in a transparent and technically understandable way, argues Anthropic.
That is exactly what is missing in this case. And that’s exactly why the dispute could go far beyond this individual case. For Europe, the dispute is above all a warning: those who do not develop the most powerful AI models themselves will at some point no longer be able to decide for themselves who is allowed to use them.