Being a local AI company means a few things to us. Amytis is being built in Scotland; we want to contribute to the technology ecosystem around us, creating useful tools and expertise here rather than treating AI as something that only arrives from somewhere else. 

We also want AI to be local in a more literal sense: running on people's own computers and reducing where possible reliance on centralised data centres.

AI sovereignty is often discussed at the level of countries, where access to computing infrastructure, models and expertise increasingly affects how dependent a country is on technology developed and operated elsewhere. 

Depending on another country for access to AI models and tools can be a vulnerable position to be in: the receiving country becomes reliant on the supplying country’s policies to ensure the AI has been built responsibly and won’t be switched off without notice. Countries without AI sovereignty may also miss out on economic growth opportunities associated with AI innovation and growth. 

The same question of control and AI sovereignty exists at every scale, from governments and universities to companies, laboratories and individuals. Control over the models we use, where they run and the data they can access matters to everyone.

Local AI (AI that runs locally on your own device) is the most practical way of asserting that control, and Amytis now supports it throughout the application.

Why should I use local AI?

The benefits are immediate. Your data can stay on your machine, and you can use your local model as much as you want. There are no API keys, usage credits or limits. You provide the model, Amytis runs it locally, and information processed by that model is not sent to an external AI provider.

A researcher may want to analyse unpublished results, a lab may hold sensitive experimental data, and a company may want AI to work with proprietary documents or internal knowledge. Keeping the model local allows that information to remain within an environment the user controls.

What can I use local AI for in Amytis?

Local models in Amytis can be used across the same kinds of workflows as cloud AI. They can respond to prompts, work with information flowing through a canvas, assist with Python and R , query databases and graphs, retrieve answers from collections of documents, and let you ideate and design complex experiments.

How can I use local AI in Amytis?

In the Settings menu, select Models on the left hand side. Scroll to the bottom of this section, beneath all the API key options you’ll see a section for Local Models.

Here, you can import a model that’s already downloaded on your device, or you can download one from HuggingFace. 

What’s next?

As local models become more capable and the hardware needed to run them becomes more widely available, we expect the range of work that can be done locally to grow considerably, and we want Amytis to help drive that progress. Our medium-term R&D will therefore focus heavily on developing our own specialist AI models, particularly for biology, designed from the outset with local use in mind.

For us, being a friendly neighborhood AI company means building powerful tools close to the people who use them, giving those people meaningful control over them, and applying that technology to biology where we think it can do some good.

Watch this space.