Context

Japan, similarly to many other countries, has had a long tradition of keeping non-lawyers out of legal practice for the protection of public interest and lawyers’ livelihood. This principal is laid out in Article 72 of the Lawyers’ Act. The provision outlaws legal practice by non-lawyers (a) for a fee, (b) in connection with disputes (cases and controversies); and (c) by providing legal services, e.g., rendering legal opinions, representing a client and providing services in arbitration or settlement, among others.

The policy was felt sound until now, when AI can provide what is suspected to be legal services. There are numerous non-layer service providers that are willing to provide AI-based legal related services today. Any negative view based on Article 72 would wipe out such market, which is estimated to be Yen 30 billion (USD20 million) in Japan today.

An acute issue has been observed in connection with the review of contracts by AI. The Ministry of Justice (the “MOJ”) had responded to an anonymous inquiry as to whether providing an AI-based contract review service for a fee would violate Article 72, stating there is indeed such a concern. There were enough debates about this view as a policy and its accompanying chilling effect, so the semi-governmental “Start-up Innovation Working Group” was launched with the participation of the MOJ and the Japan Federation of Bar Associations. The Group, after deliberation, recommended that there should be governmental guidelines on the use of AI in contract review service. In August 2023, the MOJ issued a memorandum entitled “Relationship between AI based Contract Review and Related Services and Article 72 of the Lawyers’ Act,” (the “Legal-Tech Guidelines”).

Scope of the Guidelines

• The Guidelines discuss contract review by AI only and not other AI-derived services.
• The scope being addressed is non-lawyer business operators offering, for a fee from a client, a service that is AI-based contract review.
• The user of such service is not penalized under the Lawyers’ Law, so the central issue is the culpability of an AI-based service provider. This is a large industry to form if we allow it.
• If an admitted lawyer is involved in screening the input of AI contract review, be it in a law firm or a corporate legal department, the use after such screening is put outside of the purview of this discussion and deemed legal.

What the Guidelines mean in Reality

These Legal-Tech Guidelines are primarily important for the non-lawyer operated corporate legal department buying AI-based contract review services from a vendor, not because it is penalized for violation itself (it is not), but because the contracted service is disrupted if found in violation of that law. Therefore, the discussion is about the vendor being paid from the non-lawyer corporate legal department. If the user is a law firm or a legal department manned by a lawyer, such issue does not exist. It is true there are numerous corporate legal departments that do not have a lawyer in them, so it is a real issue.

In such circumstance, how can the corporate legal department get around the non-lawyer providing legal service prohibition? The Legal-Tech Guidelines make the following points:

1. The “fee” requirement is interpreted broadly on an overall quid pro quo basis, not simply on the fee vis-à-vis the service.
2. The “dispute” requirement outlaws reviews tailored to provide resolution to a preexisting dispute. Once a dispute element is identified, a lawyer should be charged with evaluating it. The test to determine a dispute hinges on the purpose of the contract, the relationship between the parties, the history leading to the contract and background. Put differently, a contract that is to resolve a legal difference is probably a dispute under Article 72.
3. The “legal service” requirement in contract review involves providing a case-specific analysis on a contract. The permitted service must be generic in nature without reference to a specific circumstance that involves concrete and detailed facts. The service should not indicate the existence of legal risk or degree thereof, nor should it offer ways to fix problems. The typical service will involve identification of differences from a template, based on semantics alone, by quoting general annotations and case law. The service should not provide suggestions for fixes.

The descriptions in these Guidelines underestimates and already is outdated in its assessment of what an AI can do in contract review. If you literally practice the Guidelines, the service would be less helpful or efficient. The only clear answer in the Guidelines for a way-around is to hire a lawyer to review the input, which could defeat the purpose of the AI contract review. Whether this policy comes from the lawyer supremacy mentality or genuine sense of public interest is an interesting issue, but it is guaranteed that the Guidelines will need to be revisited soon.