Policies to Ensure that AI Benefits Free Access to Law

Abstract

This paper examines what policies concerning artificial intelligence (AI) applications to law will ensure the greatest benefits for the goal of free access to law. The AI systems particular those using ‘generative AI’ and those based on ‘rules as code’ require particular consideration at present. It considers two types of policies: first, Government policies adopted in relation to access to, and uses made of, legal information originating from the State; and second, the policies adopted by LIIs (legal information institutes) themselves.LII policies are illustrated mainly by AustLII (Australasian Legal Information Institute) practices.

The scope of the data available via AustLII is explained, and that it is vital for AustLII to maintain the uniqueness of what it provides, as this comprehensiveness may give advantages for the development of AI applications to law, and is the key to maintaining AustLII's funding. Such a corpus, we argue, is essential for the development of generative AI applications on Australian law, and may be for the development of large scale 'rules as code'.

We argue that a new era of monopolies over legal information could occur through government monopolies over access to use of the fine-tuned LLM or the RaC that it has developed. Government Competition in the use of publicly-funded fine-tuning of LLMs and RaC must be on a level playing field, with everyone having equal opportunity to develop high quality applications using the underlying LLM or RaC.

In  relation to LII policies, 'free access to law' in future will have to include AI applications or free access will again become inferior to both government and commercial publishing.
We explain the various ways by which AI may facilitate creation of new types of monopolies and oligopolies. However, LIIs are not defenceless, and may have unique advantages, provided the cost impediments of developing LLMs or Law as Code can be overcome.

A crucial LII policy is that LIIs are not repositories. It is implied by the FALM Declaration, but not explicit, that LIIs have no obligation to provide their data to others for republication.  Copyright law, cannot be relied upon to protect data held by LIIs, so the best answer to mass copying by LLM developers may well remain the use of the Robots Exclusion Protocol (if it is observed), and technical blocking of rogue domains.

AustLII’s conclusion, based on its experience, is that every LII needs to develop its own set of policies, but AustLII recommends adoption of a very restrictive policies, particularly while generative AI is still developing.

Keywords: free access to law, artificial intelligence, AI, monopolies, AustLII, legal information institutes