Specialized AI Outperforms General Models in Tax Tests
Nylon’s research platform delivers higher accuracy than standard AI by querying a defined set of trusted legal sources.
Updated on Oct. 29, 2026 in Artificial Intelligence
Live Poll
Would you trust an AI tool to perform professional research for tax or legal advice?
Nylon, formerly known as Law Cyborg, provides AI-driven tax and legal research software for over 2,000 professional firms. The company has demonstrated higher performance on CPA taxation benchmarks compared to leading general-purpose AI models.
Why it matters
Professional advisers require verified in-text citations and high accuracy for advice they sign, tasks where specialized AI systems currently outperform general-purpose models.
In a benchmark test of 91 questions from the CPA Program's Australia Taxation module, Nylon achieved a 4.7 score out of 5, outperforming Claude (4.32) and ChatGPT (4.28). The system maintains performance by querying a defined set of trusted sources combined with a citator process.
The players
Nylon
An Auckland-based software firm that provides specialized AI-driven technical tax and legal research tools for accounting and law practices.
Claude
A general-purpose large language model developed by Anthropic that was included in the comparative benchmark test.
ChatGPT
A general-purpose generative AI platform developed by OpenAI that was included in the comparative benchmark test.
The details
Nylon functions by anchoring its responses to specific legislation, case law, and commentary. The system provides in-text citations and a confidence score for its outputs, ensuring that users can verify the information against the underlying source material. Additionally, the company offers a meeting bot that monitors real-time discussions for tax exposure or compliance risks.
Timeline
Four years ago the company was founded in Auckland.
The Tech Race
Nylon is positioning its specialized legal and tax AI against broader general-purpose models like Claude and ChatGPT. By focusing on professional taxation benchmarks, the firm seeks to prove that vertical-specific data architectures remain superior for regulated advisory fields.
Accounting and law firms currently using the software can access automated tax and legal research integrated into existing workflows. The meeting bot feature provides immediate real-time risk assessment during client engagements for subscribers.
The takeaway
Nylon proves that specialized training on verified legal and tax databases yields higher accuracy than general-purpose LLMs for professional certification tests. Users should watch for whether the company expands its benchmark testing to include other international regulatory jurisdictions.
Further reading
Explore how Artificial Intelligence is reshaping professional research workflows and software accuracy.
Source note: This article includes information reported by AccountantsDaily.
Live Poll
Would you trust an AI tool to perform professional research for tax or legal advice?
