While on his recent visit to Ireland Donald Trump stated, “We're the most sophisticated country in the world, and frankly I want to keep it that way because whoever wins AI, wins”. His comments were triggered by the barrage of media noise that had been created by the resignations of two researchers from OpenAI and Anthropic, both of whom raised major concerns about the safety of AI models.

And they're not the only ones, with some of these concerns coming from surprising sources. The prime examples being quotes from Satya Nadella, CEO of Microsoft, and Alex Karp, CEO of Palantir, both of whom highlighted the risks that companies face when they expose too much of their data to the AI model providers, and in doing so provide them with all the information they need to become direct competitors. And this isn't just idle theorising, Ole Lehmann calls out a significant list of examples where this has happened. The most blatant example he quotes is where Figma partnered with Anthropic on AI design tools. Then Anthropic's product chief quit Figma's board, and 3 days later Anthropic launched Claude Design to compete with Figma. CEO Dylan Field said Anthropic was “not consistently candid”.

Other concerns have been raised about the rising cost of the frontier models and about the concentration of power in a small number of companies and individuals.

But what is the impact on European businesses and are there any specific issues we should be thinking about? A number of businesses are starting to believe there are. Models have been held back from release to UK testers, raising concerns about the ability of UK regulators to have a voice in discussions about the future direction of these models. And US labs have been prevented from releasing the latest versions of models to Europe. This was in a specific security related scenario, but it creates a further concern around the US being able to use model availability as leverage in wider discussions, or to disadvantage foreign competition.

So, what's the solution? Luckily, there is a simple solution that addresses all of these concerns, and its a solution that ties in with widely accepted best practice within the technology sector. That practice is to make sure that technology strategy flows directly from broader business goals. So, to focus on a technology such as the latest frontier LLM, and then to look for ways to use that technology within your organisation, is often an expensive mistake. Rather, a better approach is to define organisational goals (eg “increase direct sales to account for 40% of revenue”) and then to align the technology strategy with the delivery of those goals. This will usually result in a much more focussed trechnology strategy, and in the world of AI that often translates to much narrower requirements for AI models.

And this is where a strategic business led approach converges with a solution to the concerns raised about AI. If by focussing on a clear business goal you can define much narrower requirements for your AI models, then it becomes possible to select smaller, more focused models as opposed to the frontier LLMs. This has many advantages:

• Smaller models are often more efficient, and so can be run on smaller, cheaper hardware. This makes self-hosting possible, eliminating the seepage of IP to third parties. Even if hosted solutions are used, the service providers will not be seeing the entire process, and will not be benefiting from the over-sharing that's been seen with larger LLMs.
• Smaller models don't require the levels of investment that we've seen in frontier LLMs, so we are seeing these models getting cheaper whilst the cost of frontier models continues to rise.
• Smaller models are widely available, from a large number of suppliers, which negates the concerns around concentrating power in a small number of companies and individuals.
• And there are many European suppliers, which negates the risk of European businesses being overly dependent on US suppliers.

So, what European models are available? Europe has a rapidly growing AI ecosystem, with companies such as Mistral, Aleph Alpha and LightOn leading the charge. Mistral produces a number of models including Mistral Large 3, an LLM which competes with the large US frontier models. But generally, European labs produce smaller application specific models that excel in areas such as sovereignty, governance and GDPR compliance.

And the list of European labs is growing. An interesting example is Cosine, a UK based lab that has grown out of the UK Sovereign AI initiative. Cosine has released Genie, its coding assistant, and has big plans to release Lumen, a frontier scale LLM, by the end of 2026.