Key numbers
2.6GB
Raw Israeli case law
The Problem
Hebrew jurisprudence is hard to search.
Israeli civil law is governed by Knesset legislation, Supreme Court precedent, and labor court rulings — most of it in Hebrew. Generic AI tools (ChatGPT, Gemini, Claude) frequently hallucinate citations when asked Hebrew legal questions, inventing case names that sound plausible but don't exist.
This is a real problem with material consequences: an Israeli citizen who acts on a hallucinated precedent can lose money, file an invalid claim, or miss a critical deadline.
The Methodology
Retrieval, not generation.
legal-eye uses a verbatim retrieval architecture. When you ask a question in Hebrew, the system:
- Classifies the legal domain (employment, contracts, consumer, tenancy)
- Identifies applicable doctrines (good faith, material breach, etc.)
- Retrieves the most relevant passages from a 2.6GB corpus of Israeli rulings + statutes
- Returns those passages verbatim — never paraphrased, never invented
Word-for-word citations
Every quote you see is real text from the corpus. Click to verify the source.
Zero data leakage
Your question never leaves Israeli infrastructure. No OpenAI / Anthropic / Google receives a copy.
No LLM text generation
We don't use ChatGPT-style models to write answers. The system can't hallucinate what it doesn't generate.
No "AI lawyer"
For real personal counsel, we match you with a bar-verified attorney — never replace one.
Why it matters
For Israelis, by an Israeli lawyer-engineer.
legal-eye was founded by Avri Barzel, a licensed Israeli attorney and software engineer, after observing the first wave of "Hebrew legal AI" tools confidently citing fabricated precedents. The platform launched in 2026 with the explicit commitment to never use a third-party LLM for answer generation.
Today, the platform serves Hebrew-speaking citizens across four primary domains (employment, contracts, consumer, tenancy) and matches users with verified Israeli attorneys within 24 hours. The full architecture is open about its limitations: when there's no good answer in the corpus, the system says so — it doesn't make one up.
A note on the language
The user-facing product is Hebrew-only. This is deliberate — Israeli law is in Hebrew, and a Hebrew-native experience is essential for accurate results. This English page exists for press, international observers, expats, and partners. If you need to use the actual product, you'll need to switch to the Hebrew side.
המוצר עצמו בעברית: legal-eye.1bigfam.com
Get in touch
Contact channels
Resources
For deeper exploration
- The Corpus — Full transparency: what's indexed, how it's updated, methodology
- Technical methodology — Deep-dive on retrieval architecture (Hebrew, for devs/researchers)
- How it works — 4-step user journey, JSON-LD HowTo schema (in Hebrew)
- Cornerstone doctrines — 6 foundational Israeli civil law doctrines explained
- 78 years of Israeli civil law — Timeline of major legislation 1948-2026
- Security & privacy — 7 pillars + compliance status
- Public roadmap — What's live, in progress, planned, and the long-term vision
- Quality eval — Weekly public benchmark, PASS/WEAK/FAIL per question
- Press kit — Logos, boilerplate copy, founder bio
- Hugging Face Space — Backend infrastructure (public)
Try it (Hebrew)
The product itself is Hebrew-only. If you read Hebrew, ask a real legal question and see verbatim citations.
לכניסה ←