Author: heise.de
Executive Summary
Journalist Martin Gerhard Loschwitz criticizes the restrictive stance of German organizations toward Large Language Models (LLMs) such as ChatGPT and Claude. Many German companies and authorities prohibit or block the use of these AI models citing reasons such as data protection concerns and unresolved copyright issues. Loschwitz argues that ignoring and banning these technologies offer no solutions but instead create competitive disadvantages. Instead, he advocates for a societal debate on AI deployment and the social consequences of the technology.
People
- Martin Gerhard Loschwitz (freelance journalist)
Topics
- Artificial Intelligence and Large Language Models
- German economic policy and innovation
- Digital sovereignty of Europe
- Labor market impacts of AI
Clarus Lead
Germany's defensive stance toward AI technologies exacerbates an existing innovation deficit and increases digital dependence on US corporations – a strategic risk made particularly clear by export restrictions such as Trump's blockade of Anthropic's Fable 5 and Mythos 5 models. While Ukraine successfully deploys AI-powered drones, European organizations miss both economic efficiency gains and opportunities to shape regulation of this key technology through preventive bans. Time for European countermeasures is running out.
Detailed Summary
Loschwitz first distinguishes between marketing rhetoric and technical reality: LLMs are not true artificial intelligence, but rather "sophisticated statistics" in the language domain. This conceptual confusion leads to inflated expectations and diffuse fears. Nevertheless, the author identifies concrete added value – not only economically but also practically. A bookkeeper in Berlin must manually sort three tax categories from a taxi receipt via Free Now, work that an AI could complete in seconds. A skilled developer with domain knowledge could write code with Claude in a few hours that would take a human weeks.
The most prominent example comes from Ukraine: For months, the country has used AI-controlled low-cost drones to intercept Russian Shahed drones that cost ten times as much. The return on investment is unprecedented – and substantial: "Every Russian projectile that does not hit Ukraine cannot kill anyone there." European military and defense officials are already aligning themselves with these solutions.
The author warns against two errors: First, Germany binds itself through bans to "a massive competitive disadvantage entirely unnecessarily." Second, categorical rejection blocks a necessary societal debate. Massive job losses are foreseeable – accountants, programmers, lawyers are considered at risk. These social impacts require proactive solutions, not avoidance. Third: when European companies must necessarily use US systems due to lack of domestic alternatives, this reinforces digital dependence. Trump's export embargo against Anthropic demonstrates: European digital sovereignty remains a phantom without independent AI competence.
Key Statements
- Bans instead of shaping: German rejection of LLMs creates no security, only dependence and lost innovation time.
- Practical utility proves itself: AI demonstrably increases productivity in everyday tasks and enables military efficiency (Ukraine).
- Societal debate necessary: Job losses caused by AI require anticipation and compensation mechanisms, not denial.
- European capacity for action endangered: Without its own AI infrastructure, Europe becomes technologically dependent on the US.
Critical Questions
Evidence: Is the claim "well-trained LLMs solve everyday tasks better and faster" based on systematic comparative studies or on the author's individual observations (e.g., the Claude experience "for fun")?
Conflicts of Interest: The author works as a journalist covering technology topics; could positive personal AI experience lead to overweighting of advantages?
Causality: Do German organizations become less innovative because of bans, or do they ban LLMs precisely because they are already averse to innovation – i.e., a symptom, not a cause?
Feasibility: How should "societal debates" on AI job losses be concretely designed and operationalized politically, given the rapid pace of technological change?
Counter-hypothesis: Could temporary usage bans not also serve as a protective space to establish regulatory frameworks before widespread adoption – rather than having to improvise afterward?
Data Quality Ukraine Example: Is the claim about Ukrainian drone efficiency based on verified military sources or on publicly circulating anecdotes? What is the verification status?
Sources
Primary Source: Commentary: AI – why ignoring and banning don't help – heise online
Verification Status: ✓ 2025
This text was created with the support of an AI model. Editorial responsibility: clarus.news | Fact-check: 2025