Sydney barrister Laina Chan saw a gap in the market for AI legal tools and sought to fill it — while still continuing her work at the bar.
She found that AI tools relied on post-trained language models that draw on a combination of model memory and secondary materials to construct answers. That’s a problem because these tools cannot reliably explain relevance or produce defensible reasoning grounded in authority.
So Chan, who has a background in science, law and pure mathematics, took the initiative. She built a legal-AI reasoning system built from the first principles of law.
Her platform, MiAI Law, applies artificial intelligence directly to case law, legislation and contracts, reasoning step-by-step from source to conclusion in a way that mirrors how lawyers think. Unlike conventional legal AI tools that primarily retrieve or summarise information, MiAI Law constructs structured, auditable analyses in which each conclusion is transparently linked back to primary legal authorities.
The system is designed to go beyond research and drafting by providing a verification layer for legal work. Its tools, including LawCheck and AppealCheck, perform a range of tasks, such as auditing legal propositions, flagging misstated principles or hallucinated cases, and identifying potential appealable errors for human review. Built to reduce risk and improve confidence in high-stakes matters, MiAI Law automates foundational legal tasks while preserving the lawyer’s role in judgment and strategy.
“We start from the building blocks of law itself – cases and legislation,” Chan told Women’s Agenda.
“We extract the controlling rules, anchor them to the precise paragraphs where the court states them, and build the analysis from that foundation of legal rules.
“Every proposition in the MiAI Law research reports is hyperlinked back to the primary source with pinpoint references. During answer generation, the language model is tightly constrained to generate a reasoned report from those inputs. Instead of asking a model to recall or infer the law, we require it to reason from the law.”
Chan spent years building the platform, reading technical papers on language models and learning how a language model behaves.
“Language models are very slippery beasts,” she said. “By that, I mean that language models are trained to be useful and to always answer a question regardless of whether there is a basis (or context) for it. This desire to always produce an answer is essentially what leads to hallucinations.”
“I am supported by my head of research, my in-house counsel, and ten developers who build and scale my legal reasoning architecture.”
With her busy schedule as a barrister, I asked her how she managed to build her own AI technology.
“I still practise, but selectively,” she revealed. “I focus on advisory work, which aligns closely with what we are building.”
Chan said she wants to focus on MiAI Law.
“At its core, it’s a risk management tool. Legal work rarely fails because the law is unavailable. It fails when something material is missed – a controlling rule, a limiting authority, or a defect in how an argument is structured.”
MiAI Law is expected to become part of the standard working environment for lawyers, not as a standalone tool, but as a verification and reasoning layer integrated into drafting and decision-making processes. Its core purpose is to help lawyers produce the highest quality work within the constraints of time, resources, and client budgets. To support this, the company is focused on embedding its capabilities directly into legal workflows. Through its partnership with Microsoft, MiAI Law has developed Microsoft Word plugins for contract and pleadings review and has also agreed to operate as an agent within Copilot.
“If lawyers embrace this way of working, then clients will be well served,” Chan said. “We see MiAI Law levelling the playing field by making all law and material risks easily discoverable. All lawyers will be empowered to deliver work that has been built on a foundation of legal principles anchored in cases and legislation with less left to chance or memory.”