With a boost and the will to shape — how politicians can shape generative AI

With the release of ChatGPT in November 2022, a global AI boom has begun, and new tools for generating text, speech, images, and videos have been introduced quickly. It is hard to imagine business, science and the general public without the topic of generative AI. According to a representative survey by the DIHK, two-thirds of companies already used AI in 2023, almost twice as many as in 2022 (DIHK 2024). According to a survey by the Boston Consulting Group (BCG 2024), generative AI, in particular, is one of the top three topics on the minds of CEOs. In a representative bidt study (bidt 2023), 37 per cent of internet users see generative AI as a technology that will make life easier in the future.

In German politics, however, the topic only seems to be slowly gaining traction. For a long time, the German government talked about regulation and the EU AI Act, while little was heard about its position on the technology — or what role Germany should play in the global AI industry and research in the future.

The federal government urgently needs to resolve this major disconnect between the very differentiated debate in science and industry about the opportunities and risks of generative AI, and its role as a silent spectator to date if it still wants to play a formative role in Europe and globally. As rapidly as generative AI is developing, politicians will only be able to catch up on their lack of knowledge and experience if industry and science help and give it a boost.

Such support would have the greatest effect on all sides in three respects:

1. The administration itself must become a Gen-AI professional

Only those who deal with generative AI daily can understand it and regulate or promote it sensibly. There are initial approaches and positive voices in the federal government. Still, compared to other countries, the federal administration is far behind in terms of use cases that make its work easier using ChatGPT and the like. There are many easy-to-implement use cases, such as the automated summarisation of documents. The best bridges for administering such tests are the data laboratories in all federal ministries. The federal government can benefit greatly from the experience of the companies when testing on a large scale. These, in turn, benefit from realistic regulation based on their practical knowledge.

2. Making risk management quickly practicable in line with the AI Act requirements

Speaking of realistic regulation, everyone working with the technology knows that generative AI entails risks. There are already much-discussed ethical and compliance issues arising from adopting the EU AI Act and many laws in other countries worldwide. In addition, the cybersecurity risks of large-scale language models and their quality deficits, such as hallucinations, are becoming increasingly apparent. Following the AI Act decision, its provisions on risk management are now being converted into ISO standards, but this may take years. Even simple guidelines could generate a great deal of innovative power. For example, the Federal Financial Supervisory Authority (BaFin) could sit down with academia and the banks to jointly clarify specific questions on the governance of high-risk use cases in the banking sector in a co-creation process. Such questions will remain unresolved in the ISO process and are causing an innovation backlog in the financial sector.

3. Appointing visible contact partners and pooling national strengths with convening power

Joe Biden has Kamala Harris and a visible and often prominent AI commissioner in almost every ministry; Emmanuel Macron has former Harvard professor Philippe Aghion and mathematician Cédric Villani — but the German government does not yet have a clear or prominent face for AI who drives the topic nationally and internationally and is a contact person. It doesn’t have to be an AI commissioner, but there needs to be convening power and political attention so that the knowledge and experience from companies and universities can come out of the silos and be pooled more strongly, and risks can be recognised and averted together. If no solution can be found regarding party politics, the German government could also follow the path of France and appoint a person from the scientific community or a CEO. The window of opportunity to help shape the future with generative AI on a global political level is gradually closing. Other countries and economic areas are becoming more active week by week. It would be good if the German government did not miss its chance.

---