The combination of artificial intelligence and law has ceased to be a future possibility and has become a reality that is transforming both the practice of law and the functioning of the Administration of Justice.
The automation of tasks, the preparation of procedural documents, jurisprudential research or the analysis of legal documentation promise to improve efficiency and reduce the costs of access to justice. However, this transformation raises questions that transcend the technological dimension and directly affect essential principles of the legal system, such as legal certainty, procedural good faith, effective judicial protection or professional responsibility.
From a legal perspective, one of the main challenges lies in determining to what extent an activity that requires regulatory interpretation, critical reasoning and assessment of particular circumstances can be delegated to algorithmic systems.
The issue becomes especially relevant when the responses generated by these systems present an appearance of legal rigor that does not always correspond to reality, transferring the risks derived from their use to the professional and, potentially, to the judicial body itself.
A paradigmatic example is found in the recent ruling of the Supreme Court, Third Chamber, Contentious-Administrative, Section 4, Sentence 803/2026 of June 29, 2026, Rec. 2298/2025. Thus, in said procedure the High Court detected in a procedural document jurisprudential references whose pronouncements did not correspond to the real content of the cited resolutions. The Court found that the fragments in quotation marks did not coincide literally or substantially with the sentences invoked, which motivated the referral of the matter to the General Council of Spanish Lawyers to assess possible professional responsibilities.
This assumption demonstrates one of the best-known risks of generative artificial intelligence, the so-called hallucinations, consisting of the production of apparently coherent and credible information, but lacking correspondence with real sources. In the legal field, its consequences are especially relevant, since a non-existent jurisprudential reference does not simply constitute a technical inaccuracy, but can compromise the foundation of a procedural claim and alter the conditions under which the judicial debate must take place.
The recent resolutions of the Spanish courts introduce a particularly relevant reflection: the availability of technological tools does not reduce the standard of diligence required of the professional, but can contribute to specifying it. The use of artificial intelligence is not prohibited, but it does not allow the responsibility derived from the content of the writings presented to the courts to be shifted to the algorithm. Human supervision thus ceases to be a recommendation of good practices and becomes a requirement linked to deontological duties and the principle of procedural good faith.
Artificial Intelligence and law: some questionable results
The technological transformation of the legal sector is also offering results that question traditional models for providing legal services. In the United Kingdom, the Garfield AI platform, authorized by the competent professional regulator, managed to prepare the necessary documentation for a legal claim of approximately £7,000, at a cost of close to £400 for the plaintiff. The procedure ended with a favorable resolution, although representation at trial was carried out by a real lawyer.
Thus, in light of the above, we must ask ourselves the following questions: when does a legal information tool become true legal advice? Who is responsible for the damages derived from their recommendations? And how far is the developer’s responsibility when the user uses the system to make procedural decisions?
These questions connect with a broader problem of attribution of responsibility in complex technological ecosystems, where developers, suppliers, professionals and end users intervene. The intervention of multiple actors can make it difficult to identify who effectively controls the risk, especially when systems generate probabilistic results and their internal processes are not fully transparent.
On the other hand, the digitalization of the legal and business environment is also redefining the limits of privacy and the use of information obtained through technological tools.
The regulations of the General Data Protection Regulation, the requirements of professional confidentiality and the European framework for regulating artificial intelligence must be highlighted, since all of this is becoming increasingly important. The use of algorithmic systems in legal environments requires assessing not only their effectiveness, but also the legality of the information processing, the purposes pursued, the proportionality of the operations carried out and the guarantees of human supervision. Furthermore, certain artificial intelligence applications linked to the Administration of Justice may be subject to obligations reinforced by their regulatory consideration as high-risk systems.
The underlying question is not, therefore, determining whether artificial intelligence should be incorporated into the Law, but rather establishing under what conditions it can do so without weakening the guarantees that justify the very existence of the legal system. Automation can help democratize access to certain services, improve document management and free up professional resources for more complex tasks, but its usefulness does not eliminate the need to preserve the veracity of sources, the independence of legal criteria and the possibility of demanding responsibilities.
In short, recent judicial pronouncements show that artificial intelligence is ceasing to be just an auxiliary tool and becoming a factor that conditions the way the Law is exercised, interpreted and applied. Its integration requires evolving from trust based on technological efficiency towards a governance model based on verifiability, transparency, security and professional responsibility.
Because, ultimately, the real challenge is not getting a machine to be able to make a convincing legal argument, but ensuring that that argument is correct, verifiable and compatible with people’s rights. Justice can rely on algorithms, but it cannot delegate to them the guarantees that define it.
