Technological spheres of influence have existed since the emergence of the internet in the latter half of the 20th century. The internet allows for instant access to a system on the other side of the planet and for the ability to change code remotely. This was not possible during previous ages, such as the time when the telegraph was the major mode of communication. The telegraph was used to communicate but not to directly control technology far away. A weaving machine for instance was controlled by a punched card. There was no way to corrupt a punched card in Germany from abroad at the time. As soon as digitization and the internet took root a weaving machine that is connected to the internet can be corrupted to malfunction (Clapson, 2020).
Artificial intelligence adds the attribute of reasoning to an external actor’s influence. For instance, before the age of AI, influence was gained by inserting backdoors into a system to undermine confidentiality, integrity and availability of systems. This would allow hegemons to siphon off data or make technology malfunction. But control was dumb. Machines followed instructions with limited impact. A nuclear enrichment facility would start spinning its turbines faster when told to do so to destroy the turbines themselves. In the age of AI all this can still be done, but the machines that carry out this control have changed. Control has become intelligent. In the age of AI, machines have much more autonomy and leeway to exert control on behalf of companies and countries. DeepSeek for instance writes erroneous code for engineering projects that concern Tibet (CrowdStrike, 2025). DeepSeek thereby puts certain users at a disadvantage. In this case influence works like this. In a first instance, the Chinese Communist Party (CCP) instructs companies to write code that aligns with values of the CCP (Sprick, 2025). In a second step DeepSeek code, which is aligned with CCP values, would carry out this command wherever and however it saw fit. On the U.S. side, the CLOUD Act allows U.S. authorities to examine data held by U.S. companies no matter where the data is kept (Foitzick, 2020). In sum, the impact in the age of AI compared to before the age of AI is much more vast, subtle and lasting than in previous ages.
Please find the full policy paper by Valentin Weber here.
