# Habeas

> Habeas is an Australian legal AI research platform. It runs search-first, retrieval-grounded AI over a closed corpus of Australian primary law — over 300,000 cases and pieces of legislation from authoritative Australian sources — so that every answer is traceable to a real source and every citation can be verified. Habeas is built for Australian legal practitioners: barristers, litigators, boutique litigation firms, and in-house counsel.

Habeas exists because general-purpose AI tools fabricate legal citations with fluent confidence, and the cost of a confident wrong answer in legal practice is measured in professional obligations and client outcomes. Habeas is architected around verification rather than generation: it retrieves from real authorities before it reasons, and links every output back to its source.

Core products:
- Search Engine: semantic search across Australian case law and legislation, returning results pinpointed to paragraph level with traceable citations.
- Document Stores: secure, matter-specific repositories that let practitioners reason over their own materials (pleadings, affidavits, evidence) without exposing data to third-party models.
- Research Assistants: configurable agentic research tasks that return structured, cited analysis rather than raw search hits.

Key facts:
- Jurisdiction: Australian law only (Commonwealth and state), not US/UK-centric.
- Architecture: retrieval-augmented generation over a closed, legitimate Australian legal dataset; designed to minimise hallucination.
- Citations: every output links to primary source material, supporting practitioners' verification obligations under court AI guidance (e.g. Federal Court GPN-AI, NSW SC Gen 23).
- Pricing: Sole Practitioner and Unlimited tiers, plus Enterprise. 14-day free trial available.

## Flagship research

- [The Dangerous Mirage: An Open Benchmark of Citation Accuracy in Australian Legal AI](https://www.habeas.ai/blog/citation-accuracy-benchmark-australian-legal-ai): An open-source, reproducible benchmark measuring how often legal AI tools (Habeas, GPT-5, Claude) invent or misattribute Australian legal citations. Distinguishes two failure modes: existence (the case isn't real) and support (the case is real but doesn't stand for the cited proposition). MIT-licensed, full methodology and raw responses published.

## Use cases and workflows

- [Where Habeas Fits in a Barrister's Day](https://www.habeas.ai/blog/where-habeas-fits-in-a-barristers-day-five-workflow-touchpoints-that-save-hours): Five workflow touchpoints where Habeas saves barristers hours — brief review, authority gathering, drafting, case-law monitoring, trial prep.
- [Why General Counsel Are Turning to Habeas](https://www.habeas.ai/blog/legal-ai-generalcounsel): How in-house counsel use Habeas as a first-line legal intelligence layer to triage risk, run compliance checks, and reduce external counsel spend.

## Perspective and positioning

- [The Hidden Tax on Legal Expertise](https://www.habeas.ai/blog/the-hidden-tax-on-legal-expertise): On the gap between keyword-search skill and legal-analysis skill, and why semantic search changes where a practitioner's effort begins.
- [What Defines 'Good Legal Research'?](https://www.habeas.ai/blog/what-defines-good-legal-research-and-why-is-this-so-important-for-ai-legaltech): A working definition — comprehensive, judicious, defensible — and how any AI research tool should be judged against it.

## On the legal AI landscape

- [Anthropic's New Legal AI Plugin: Does It Change Things?](https://www.habeas.ai/blog/anthropics-new-legal-ai-plugin---does-it-change-things): Why a general-purpose contract workflow and a purpose-built Australian research engine are different tools that coexist.
- [Federal Court GPN-AI: What It Means for Australian Legal Practice](https://www.habeas.ai/blog/federal-court-gpn-ai-what-it-means-for-australian-legal-practice): The Federal Court's binding AI practice note and its verification and disclosure obligations.

## Contact

- Demo and sign-up: https://www.habeas.ai
- Blog: https://www.habeas.ai/blog