All news

Britain's AI review targets data rights and spin-out terms first

A classical pale stone hall model joined to a plain graphite cube by a narrow bridge with a cobalt line

The UK review asks for text and data mining by default and common spin-out terms. Both cut what an AI company pays to exist before it earns a pound.

Britain should make the right to read published research also the right to mine it with software, an independent review for the government concluded on 15 October. Led by Dame Wendy Hall and Jérôme Pesenti, the review sets out 18 recommendations, from data trusts and 200 extra PhD places in AI to common terms for university spin-outs, and puts the prize at £630 billion added to the UK economy by 2035.

The PhD numbers matter least to investors. The recommendations that matter to capital lower the price of two inputs a young AI company cannot simply buy: the right to train on data it can read, and a clean claim to the ideas its founders developed in a university lab. Both costs land at formation, when every point of equity and every month of legal work weighs most, and both decide whether Britain's research turns into companies it keeps.

Current law already allows text and data mining, but only for non-commercial research, under an exception in force since June 2014. A startup that wants to train a model on published papers sits outside it. The review asks for mining by default on published research where the result does not substitute for the original, and for publicly funded research to release its data in machine-readable form. Data trusts, agreed frameworks for sharing data securely between those who hold it and those who use it, would carry that logic beyond research.

On spin-outs the language is blunt. The review calls the allocation of equity between universities and founders complex and contentious, and asks universities to adopt clear and, where possible, common policies for licensing IP and forming companies. That matters because Britain produces AI companies faster than it keeps them. More than 200 startups and small firms build AI products in the country, and US technology companies bought Evi, DeepMind, VocalIQ, SwiftKey and Magic Pony between 2012 and 2016. Cheaper formation lets more companies grow to a size a European investor can price before a buyer arrives.

Founders leaving a UK lab need not wait. The IP licence and the university's stake belong in writing before the first priced round, because investors discount an unresolved claim on the core code. Fanaura asks the same question of any machine learning company built on university research, and it comes first in how we work with founders: who owns the models, the code and the right to use the data, and on what terms.

A review, however, is not a law. Nothing in it binds universities or ministers, the mining change needs copyright reform, and the money for PhD places has yet to be found. The business secretary has promised to work with the sector in the coming months on a sector deal for AI. Whether that deal funds the PhD places, pilots a data trust and moves on text and data mining will show how much of the review becomes policy. Britain has the research and the founders. The review goes after the terms that turn one into the other, and cheap terms at formation are where early capital earns its return.

For founders

Spinning a machine learning company out of a university lab? Send us the project.

Fanaura invests its own capital and the capital of its limited partners (LPs), usually through a separate SPV for each deal. If we see a fit, we get in touch.

Apply in five minutes
Subscribe

Get the news by email

One short note each weekday. No marketing, and you can unsubscribe at any time.