Using Claude Code to analyse medical MRI results

The big AI platforms are starting to draw lines around who gets to use their models, and if your business runs on AI tools, this is the moment to pay attention.

The Cracks in the "Open" AI Market

Google has moved to restrict Meta's access to its Gemini AI models. That is not a minor licensing squabble. That is two of the largest technology companies in the world deciding that AI capability is now a competitive asset worth protecting, not a service to be sold to all comers. The era of "AI as a utility, available to anyone with a credit card" is quietly closing.

At the same time, geopolitics is reshaping where AI development happens at all. Austria is actively lobbying the European Union to have Anthropic (the company behind Claude) establish a presence there, following US government restrictions that have complicated how AI companies operate internationally. The EU wants sovereign access to frontier AI. The US is treating model access as a national interest. These are not abstract policy debates; they are structural shifts that will determine which tools you can use, where your data goes, and what alternatives exist if your preferred AI provider suddenly becomes unavailable in your region.

What This Actually Means For the Industry

The competitive logic here is straightforward. Google does not want Meta building products on top of Gemini that compete with Google's own products. Why would they? The same dynamic exists across the board. When AI capability is your core differentiator, you stop lending it to rivals.

For smaller players, this creates an uncomfortable reality. The AI tools you rely on are sitting on infrastructure owned by companies with their own commercial interests. Those interests do not always align with yours. Access can be restricted, repriced, or restructured at any point, for reasons that have nothing to do with you.

The Austria/Anthropic story is the other edge of the same blade. Regulatory and political pressure is beginning to fragment the AI landscape by geography. A tool that works seamlessly today might face access restrictions, data residency requirements, or outright blocks depending on where your business operates or where your clients are based.

What This Means If You Run a Business

If you have built your workflows around a single AI provider, you are more exposed than you probably realise. This is not scaremongering. It is the same logic as not hosting your entire business on one platform with no backups. Concentration risk is real, and the AI market is consolidating in ways that increase it.

The practical concern for freelancers and small businesses is less about geopolitics and more about continuity. If the tool you depend on becomes unavailable, repriced significantly, or degraded because the company behind it decided to prioritise enterprise clients, what is your plan? Most people do not have one.

We have seen this pattern before with cloud services, payment processors, and social media reach. AI is following the same trajectory, just faster.

What To Do About It

1. 1.Audit your AI dependencies. List every AI tool your business uses and identify which are business-critical. If one disappeared tomorrow, what breaks?

1. 1.Test at least one alternative for each critical tool. You do not need to switch, but you should know what you would switch to. Familiarity with a backup takes ten minutes now and saves significant pain later.

1. 1.Check the data residency terms of any AI tool you use with client information. With EU regulatory pressure increasing, the geographic location of data processing is becoming a real compliance consideration, not just a footnote.

1. 1.Follow the provider landscape, not just the product features. The business relationships between AI companies matter. When Google restricts Meta, downstream effects ripple through the ecosystem. Staying aware of this keeps you from being blindsided.

1. 1.Talk to whoever handles your tech about a contingency plan. Even a rough one. "We would switch to X" is infinitely better than nothing.