Beyond AI: Why credible legal information matters more than ever
Artificial intelligence is changing the way legal professionals, researchers and librarians discover and work with information. Tasks that once required complex search strategies can now begin with a simple question posed in natural language, returning relevant information in seconds.
But as AI becomes increasingly integrated into legal research, the conversation is shifting from capability to confidence. The question is no longer simply what AI can do, but how trustworthy its answers are. In legal research, confidence depends on more than speed. It depends on the quality, provenance and verifiability of the information behind every response.
Why this conversation matters
Artificial intelligence is reshaping the legal research landscape, prompting important conversations about how legal professionals discover, evaluate and apply information. As a long-standing provider of South African legal information, Sabinet believes these conversations should extend beyond the capabilities of AI to the quality and integrity of the information that supports it.
This article explores the principles that underpin trustworthy AI-assisted legal research and the role they play in developing tools that support legal practitioners, researchers and librarians. These same principles have informed the development of Sabinet’s Legal Research Assistant, designed to complement established legal research workflows within the broader Sabinet Discover ecosystem.
AI is changing legal research – but trust remains essential
There is little doubt that artificial intelligence has the potential to improve legal research by helping users navigate large volumes of information more efficiently. Used responsibly, AI can accelerate discovery, identify connections between legal sources and reduce the time spent locating relevant information.
Recent research, however, suggests that AI should be viewed as a research assistant rather than a substitute for professional judgement.
A landmark study by Magesh et al. evaluating commercial AI legal research tools found that even specialist legal AI systems still require verification and professional oversight.
Similarly, researchers from Stanford University demonstrated that large language models can fabricate legal authorities or present inaccurate legal information with confidence. Their work reinforces an important principle: the value of AI depends not only on the sophistication of the technology, but also on the quality and reliability of the information it uses.
For legal professionals, this distinction is significant. Accuracy is not optional, and verified legal sources must support every conclusion.
Why source quality matters
The reliability of any AI system is determined by more than the model itself. It depends on how the underlying information is selected, maintained and governed.
In legal research, accurate information has several defining characteristics. It should be updated and relevant to the jurisdiction in which it is being applied, and traceable back to its original source.
This is particularly important in South Africa, where legislation is regularly amended, and new judgments continually shape legal interpretation. AI systems drawing on incomplete, outdated or non-jurisdictional information may produce responses that appear convincing but are unsuitable for professional use.
Why provenance and verification matter
One of the strengths of traditional legal research has always been the ability to verify information. Researchers expect to trace legal arguments back to legislation, case law or scholarly commentary and to evaluate those sources for themselves.
The same principle should apply to AI-assisted research.
Rather than presenting answers as isolated statements, AI should enable users to understand where information originated, explore the supporting material and exercise their own professional judgement.
A reliable AI legal research workflow
These principles—accurate and up-to-date legal information, transparent source attribution, verification, and support for established legal research workflows—provide the foundation for responsible AI-assisted legal research. They also underpin the design of Sabinet’s Legal Research Assistant, which complements research conducted through Sabinet Discover by helping users navigate relevant legal information more efficiently while maintaining visibility of the underlying sources.
The simplified workflow below illustrates how these principles come together in practice.
How Sabinet puts these principles into practice
Provenance and verification have guided the development of Sabinet’s Legal Research Assistant.
These principles—accurate and up-to-date legal information, transparent source attribution, verification, and support for established legal research workflows—have guided the development of Sabinet’s Legal Research Assistant. The Legal Research Assistant complements existing research conducted through Sabinet Discover by providing an AI-assisted way to explore Sabinet’s curated South African legal collections, including legislation, judgments and law journal articles. Responses are linked to the underlying source material, allowing users to verify information and continue their research with confidence.
This approach reflects a broader philosophy: AI should enhance legal research, not replace the critical thinking and professional judgement that underpin it.
“The value of AI in legal research depends not only on the sophistication of the technology, but on the quality and integrity of the information behind it.” — Sanet Vos, Head of Product, Sabinet
For legal practitioners, researchers, librarians and compliance professionals alike, the real value of AI lies not simply in faster answers, but in answers that are grounded in verified legal information.
As AI continues to evolve, that distinction may prove to be its most important feature.
Experience AI-assisted research in practice
The Legal Research Assistant complements the Sabinet Discover research environment by combining AI with curated South African legal information, transparent source attribution, and up-to-date, credible legal content.
→ Explore the Legal Research Assistant
Further reading
- Magesh et al. (2025). Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools.
- Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models (2024).