Kenneth Glad
Public institutions are moving more work into automated systems. A recent shutdown has shown how quickly access can become a political issue.
Europe’s dependence on American AI providers is under sharper scrutiny after a US order led Anthropic to block users outside the United States from its Fable 5 and Mythos 5 models.
According to TV 2 Denmark, the models were among Anthropic’s most advanced systems and had already been used by companies and public authorities in Europe.
For organizations building services around them, the shutdown was not a minor software problem. It showed that access to AI can depend on political decisions made far from the users who rely on it.
Christian Have, technical director at BullWall, a Danish cybersecurity company that develops ransomware protection and AI-based security solutions for public authorities and private businesses, told TV 2 that the shutdown changed his view of how dependable foreign AI services really are.
Public work depends on stable access
Across Europe, AI is being considered for welfare administration, cybersecurity, health support, planning and document-heavy casework.
In Denmark, the government expects AI to free at least 50 million public-sector working hours by 2035, equal to about 30,000 full-time jobs, writes TV 2.
Similar pressures exist across the EU. Aging populations, staff shortages and long case queues have made automation attractive to governments trying to maintain services without adding large numbers of employees.
A hospital using AI to support patient administration, or a municipality using it to sort applications, cannot easily replace an advanced model overnight. The same applies to tax offices, benefits agencies and national security services.
Jan Damsgaard, a professor at Copenhagen Business School, told the broadcaster:
It is a threat to Europe that the United States can cut off access to AI like this.
Europe relies on American infrastructure
The Anthropic case sits inside a larger European problem: Cloud, software and AI infrastructure are still dominated by US companies.
That reliance is not only about convenience. Powerful AI models need enormous computing capacity, specialist engineers and large private investments, areas where Europe has struggled to match the United States.
Andreas Holbak Espersen, tech industry director and head of digital policy at Dansk Industri, Denmark’s largest business and employers’ organization, told TV 2 that political intervention is no longer a theoretical risk. He said European companies need clearer rules and stronger investment conditions to compete globally:
Europe is a tech desert. We buy what others build.
Sovereignty now means infrastructure
The digitalisation minister of Denmark, Christina Egelund, acknowledges the vulnerability:
We have rather naively bought into platforms that we do not control ourselves.
Companies can reduce risk by avoiding dependence on one supplier, keeping critical systems portable and spreading data across several providers.
That will not solve the larger issue. Europe needs models, servers, investment and companies able to operate at scale.
The Anthropic shutdown has made the policy question concrete: AI can help public services work faster, but foreign-controlled systems can leave governments exposed when access changes without warning.
Sources: TV 2 Denmark
