The recent Anthropic case made me think about a risk that many small teams ignore.
The US government ordered Anthropic to stop foreign nationals from using its Fable 5 and Mythos 5 models. Anthropic then disabled both models for all customers to comply. Other Anthropic models were not affected.
For me, the lesson is bigger than AI.
When we build products, we often choose services like cloud, CDN, databases, AI APIs, or SaaS tools because they are good and easy to use. This helps us ship faster. But it also creates dependency.
Usually we think about vendor lock-in in a technical way. A company can raise prices, remove a feature, change an API, or stop supporting a service. Then migration becomes hard.
Now we also need to think about jurisdictional risk. A service can be affected by the company, but also by the government and laws of the country where the company operates.
This does not mean every US service is unsafe. It also does not mean every project needs an expensive multi-cloud architecture. But it does mean that critical projects should have an exit plan.
For important systems, we should ask simple questions:
- Can we export our data?
- Do we have tested backups?
- Can we move to another provider?
- Do we use open formats and standard tools?
- Do we know what happens if one API disappears?
This is also why Europe needs strong and competitive cloud, AI, and SaaS services. Not because US services are bad, but because choice makes us safer.
Good architecture is not only about scaling. It is also about being able to leave.