First AI law firm in The Netherlands?

What is Garfield AI?

Garfield AI is a business founded by former litigator Philip Young and quantum physicist Daniel Long. The company aims to address the estimated £6-20 billion in unpaid debts that go uncollected in the UK each year. This is largely because the cost of legal enforcement outweighs the sums owed. Garfield AI provides legal assistance to small and medium-sized businesses, as well as other law firms, by offering an AI-powered litigation assistant specifically designed to help recover unpaid debts through the English and Welsh small claims court process (for claims up to £10,000).

Clients can upload raw data, such as unpaid invoices, correspondence and contracts, and the system will transform them into court-ready documents. Garfield AI extracts key facts, assesses debtor solvency, and generates letters that comply with the applicable Civil Procedure Rules. The system is capable of calculating statutory or contractual interest and can even recommend next steps based on debtor responses, whether that means applying for a default judgment or advising on how to negotiate. At each stage of the process, the client can review, modify and approve drafts before anything is filed.

SRA Approval

In accordance with the UK Legal Service Act (LSA), regulatory requirements stipulate that a law firm must have a designated regulated solicitor who is ultimately responsible for ensuring the firm maintains high professional standards. Additionally, at least one person within the team must be a solicitor. Under SRA rules, these solicitors are responsible for all system outputs and for anything that goes wrong. In the case of Garfield, the responsibility for this role within the company lies with Philip Young, the company's co-founder. He has confirmed that he is personally reviewing every output during the initial phase. This arrangement, which ensures that someone with the appropriate qualifications is accountable for the AI's actions, most likely secured Garfield's official approval from the SRA. It also underscores that, at its core, Garfield AI remains a law firm. Clients have the opportunity to interact directly with Garfield's technology, but a human lawyer remains involved in the process to review all outputs and is ultimately responsible for the results. This clearly satisfies the requirement under the LSA, hence Garfield functions not only as an AI platform but also as a duly regulated law firm.

Prior to authorisation, the SRA collaborated closely with the founders to assess the firm's processes, ensuring that the SRA's rules and standards could be met by an AI service. They made sure that the appropriate processes are in place to quality-check the work, keep client information confidential, and safeguard against conflicts of interest. Moreover, the SRA examined the risk of AI-generated “hallucinations”, a known concern where systems may invent case law, but found this was not an issue with Garfield AI as the system is unable to generate case law. The SRA also emphasized that Garfield is not autonomous and will only take steps where the client has approved them. 

AI Law firm in the Netherlands?

The approval of Garfield AI in England and Wales has led to speculation about the potential establishment of an AI law firm in the Netherlands.

In the Netherlands, the legal profession is governed by the Dutch Lawyers Act (Advocatenwet) and the regulations of the Dutch Bar Association (Nederlandse Orde van Advocaten, NOvA). In accordance with this framework, the title of 'advocaten' (lawyers) may only be used by individuals who are registered with the NOvA. Furthermore, any entity that presents itself as a law firm must be operated either by a lawyer as a sole proprietorship or as a partnership with other lawyers, civil-law notaries, tax advisers, or patent attorneys, in accordance with Section 5 of the Dutch Lawyers Act.

This means that a purely algorithmic platform would be unable to adopt these protected titles. The Dutch regulatory framework assumes a human lawyer who is accountable under both disciplinary and civil liability regimes. Therefore, a non-human "firm" would not meet the registration stage requirements. However, Dutch regulations do not explicitly prohibit the use of AI as a tool within a law firm. Should a qualified Dutch lawyer be appointed to review and be accountable for every AI-generated document, mirroring Garfield AI's model of client-approval checkpoints and the firm complies with the practice structure requirements of the Dutch Lawyers Act, then it is possible that an AI-assisted law firm could potentially operate within the existing regulatory framework.

Dutch lawyers are bound by five core values as stipulated in the Dutch Lawyers Act: independence, impartiality, expertise, integrity, and confidentiality. Any AI law firm in the Netherlands would need to demonstrate that it can uphold these fundamental principles. As with Garfield AI, it would be essential to ensure that the appropriate processes are in place to quality-check the work, keep client information confidential and safeguard against conflicts of interest, in order to preserve these values.

The Dutch legal system could serve as a promising gateway for AI-based legal solutions comparable to Garfield AI. In particular, the Subdistrict Court (kantonrechter) is responsible for dealing with civil disputes with a value of up to €25,000. In such cases, there is no requirement for the parties involved to be represented by a lawyer. An AI-powered law firm that specialises in debt collection could be highly beneficial within this system.

Conclusion

In conclusion, although Garfield AI appears to be an autonomous AI system that streamlines and automates the debt-recovery process, it is, in fact, a fully regulated law firm under English law. The viability of a similar model in the Netherlands is dependent on a number of factors. Under current rules, the profession requires a registered lawyer to be on record, and the idea of a hybrid entity featuring both an AI platform and fully accountable lawyers does lie within the realm of possibility. Only time will tell whether Dutch regulators will follow a similar path to the SRA, but Garfield's example demonstrates that the right balance of human oversight and technological innovation can, in principle, clear the regulatory hurdles.

The authors thank Amber van Asten for her contribution to this blog post.