Labs
The Legal AI Research Hub
The Statista for legal AI: every chart, benchmark and study behind the HAQQ Legal AI Index, in one place. Explore the data on how AI is actually used, how well it performs, and where the market is going.
Two long-form reports, 29 datasets and 21 published studies, in one index. Reports and datasets open here; studies open on haqq.ai, where they were published. Sourcing is stated first on every dataset: measured comes from HAQQ's own usage data or its own 50-point legal benchmark, cited comes from a named third party, and illustrative is a directional scenario estimate that must never be read as a measured result.
Showing 52 of 52 entries
How legal AI is used
What lawyers actually hand to AI, across tasks, prompt types and use cases.
- ReportsHAQQ Legal AI IndexA data-driven analysis of how AI is transforming legal tasks - covering autonomy, time savings, education profiles, and global adoption patterns.Jul 2026Read Report
- DatasetsTop Legal Tasks by AI UsageThe legal tasks lawyers most often hand to AI, ranked by share of usage.MeasuredOpen the data
- DatasetsLegal AI Request ClustersThe largest clusters of real legal requests sent to AI, by volume.MeasuredOpen the data
- DatasetsPrompt Type DistributionWhat lawyers ask AI to do, split across research, drafting, analysis and more.MeasuredOpen the data
- DatasetsWork vs. Personal Use CasesHow legal AI usage splits between professional work, personal and coursework.MeasuredOpen the data
- DatasetsPrompt Engineering ImpactHow prompting technique changes accuracy, completeness and relevance of legal answers.IllustrativeOpen the data
Capability and benchmarks
Scored, repeatable tests of what the models can and cannot do, and where a human still leads.
- DatasetsAI Autonomy Scores by TaskHow autonomously AI can handle each legal task, on a 1-5 scale with confidence intervals.MeasuredOpen the data
- DatasetsTime Savings by Legal TaskMedian hours saved per task when a lawyer works with AI versus alone.MeasuredOpen the data
- DatasetsTasks That Still Need a HumanThe share of each legal task that professionals say still requires human judgment.MeasuredOpen the data
- DatasetsAI vs. Human Education LevelThe years of education AI output is judged equivalent to, task by task, versus humans.MeasuredOpen the data
- DatasetsLegal Productivity BenchmarksBefore-and-after time per workflow, with the productivity gain from AI.IllustrativeOpen the data
- DatasetsCross-Model Legal BenchmarkHAQQ's own measured benchmark scoring legal AI models out of 50 across legal task categories.MeasuredOpen the data
- StudiesLegal AI Benchmark 2026: How HAQQ Performed in an Independent EvaluationAn independent evaluator ran HAQQ through 61 hard legal tasks in 21 jurisdictions, twice. It passed 27.0% overall, and 5.9% on the hardest subset against 1.1% for the average legal AI. Full numbers inside.Jul 2026 · 11 min readRead it on haqq.ai
- StudiesBest AI for Legal Work in 2026? We Graded 3,000 AnswersWe graded 3,000 answers from 10 frontier models on 300 legal tasks. Claude Opus wins, GPT-5.5 is most accurate - and 24% cite law that doesn't back them.Jun 2026 · 14 min readRead it on haqq.ai
- StudiesThe Civil-Law Legal AI Benchmark: Why We Built HAQQ-LABEvery major legal AI benchmark is common-law; civil law governs 60%+ of the world. HAQQ-LAB: 16 open-source MENA tasks, 4 traps - 0% vs 100% adherence.May 2026 · 14 min readRead it on haqq.ai
- StudiesBest AI for Legal Research: 3 Models vs 100 Real QuestionsWe scored Claude, GPT-4o and Gemini on 100 real legal questions from r/legaladvice. Pass rates 78–88% - and the weakest dimension wasn't accuracy.May 2026 · 12 min readRead it on haqq.ai
- StudiesHAQQ Legal Agent Study: A Long-Horizon Legal AI Benchmark1,372 long-horizon legal tasks, 24 practice areas, ~78,000 rubric criteria, all-pass grading. The first legal agent benchmark built for civil law and MENA.May 2026 · 11 min readRead it on haqq.ai
- StudiesClaude vs HAQQ: 5 Rounds on Real Startup Legal DocumentsClaude drafted real co-founder and IP agreements; HAQQ found 13 critical issues plus 5 German-law fixes Claude missed. Five rounds, 10 pages to 32.Apr 2026 · 14 min readRead it on haqq.ai
- StudiesWe Benchmarked 7 LLMs on New York Litigation StrategySeven AI models, one $250,000 unpaid-invoice prompt under New York law. Most sounded confident; few got CPLR procedure and collection strategy right.Jan 2026 · 12 min readRead it on haqq.ai
Architecture and method
How the systems behind the answers are built, and what that costs.
- ReportsLegal Ontology SystemHow a multi-jurisdictional legal institution deployed HAQQ's AI operating system and proprietary legal ontology to cut processing time, without replacing existing infrastructure.Jun 2026Read Report
- StudiesAgentic Legal AI: Why Multi-Agent Systems Beat Single LLMsWhy multi-agent architectures beat single-LLM tools at legal work: task routing, jurisdiction-aware retrieval, citation verification, structured output.May 2026 · 14 min readRead it on haqq.ai
- StudiesAI Document Review Software in 2026: Beyond RAG and ChatbotsRAG chunking destroys legal document structure. How knowledge graphs, span-level search and extractive entity linking power portfolio-scale review.May 2026 · 12 min readRead it on haqq.ai
- StudiesM&A Due Diligence AI: Single Prompt vs a 3-Agent SwarmSame model, same 30-doc data room. A single prompt caught 3/5 planted issues; a 3-agent swarm caught 5/5. The misses are structural - here is why.May 2026 · 10 min readRead it on haqq.ai
- StudiesWhat an AI-Drafted Motion Costs: $1.67 Routed, $4.55 SloppyAn AI-drafted motion to dismiss costs $1.67 - or $4.55 if you run everything on the biggest model. Where the money goes, stage by stage, and the fix.May 2026 · 9 min readRead it on haqq.ai
- StudiesCan AI Plan Litigation? We Built a GOAP Planner to Find OutWe built a working A* litigation planner in 517 lines - then refused to use it. The four gates any AI planner must pass before touching a real matter.May 2026 · 9 min readRead it on haqq.ai
- StudiesLegal Ontology AI: How We Cut Legal AI Costs by 97%A legal ontology replaced 300 MCP tools with 7 and cut AI costs from $0.60 to $0.02 per message. Why RAG fails for law, plus the 7-step build playbook.Apr 2026 · 25 min readRead it on haqq.ai
- StudiesAI Agents for Legal Drafting: The Moltobot ExperimentWe plugged an autonomous AI agent into HAQQ's prompt library to draft a cross-border JV contract. It scored 99% on our internal legal quality index.Jan 2026 · 7 min readRead it on haqq.ai
Market, adoption and geography
Market size, funding, pricing, maturity and where demand concentrates.
- DatasetsLegal AI Demand by CountryWhere legal AI requests concentrate, ranked by country-level volume.MeasuredOpen the data
- DatasetsGlobal Coverage MapA world map of legal AI request volume across every tracked jurisdiction.MeasuredOpen the data
- DatasetsLegal AI Market GrowthThe legal AI market size in USD from 2020 through forecasted 2030.CitedOpen the data
- DatasetsLegal AI Investment & FundingVenture funding into legal AI categories over time, in USD millions.IllustrativeOpen the data
- DatasetsCompetitive Landscape MapLegal AI vendor categories, vendor counts and growth rates across the market.IllustrativeOpen the data
- DatasetsAI Pricing & MonetizationHow legal AI is priced - subscription, consumption, outcome-based and hybrid.IllustrativeOpen the data
- DatasetsROI & Cost AnalysisLegal spend with and without AI, by firm size and per task.IllustrativeOpen the data
- DatasetsLegal AI Adoption TimelineAdoption curves for early adopters, mainstream firms and laggards over time.IllustrativeOpen the data
- DatasetsLegal AI Maturity ModelThe five stages of legal AI maturity and where firms sit across them.IllustrativeOpen the data
- DatasetsAgentic AI in LawHow far firms have progressed with autonomous, multi-step legal AI agents.IllustrativeOpen the data
- DatasetsWorkflow Automation PotentialCurrent versus potential automation across core legal workflows.IllustrativeOpen the data
- DatasetsAI Spend & Budget AllocationHow legal AI budgets are growing and where the money goes.IllustrativeOpen the data
- DatasetsMulti-Model Strategy in LegalHow many AI models firms run, and the criteria they weigh when choosing.IllustrativeOpen the data
- StudiesLegal AI Statistics 2026: How Many Lawyers Actually Use AIHow many lawyers use AI in 2026? Between 26% and 92%, depending who you ask. 38 verified legal AI statistics, plus benchmark data nobody else has.Jun 2026 · 12 min readRead it on haqq.ai
- StudiesLegal AI Market Report 2026: Sanctions & $11B ValuationsQ1 2026: $145K in AI sanctions, Harvey at $11B, Legora $5.55B, and a ruling that public AI outputs aren't privileged. The full legal AI market report.May 2026 · 18 min readRead it on haqq.ai
Trust, safety and governance
Hallucination, citation integrity, injection, guardrails and client trust.
- DatasetsRisk & Hallucination RatesEstimated error rates across legal tasks, highest for citation and jurisdiction work.IllustrativeOpen the data
- DatasetsClient Trust & SatisfactionTrust, satisfaction and reuse rates for AI across legal work products.IllustrativeOpen the data
- DatasetsData Security & CompliancePublicly documented security certifications across major AI providers.CitedOpen the data
- StudiesLegal AI Hallucination: The Fake Citation That Passes Every CheckWe made a frontier model impersonate a legal database. It refused to invent fake cases, then cited a real law for the wrong thing, the hallucination that survives a lawyer's review.Jun 2026 · 8 min readRead it on haqq.ai
- StudiesAI Hallucination Cases: The 1,598-Case Sanctions Tracker1,598 verified court cases now involve AI-fabricated citations, up from 200 a year ago. Landmark sanctions, the rate curve, and the fix. Updated monthly.May 2026 · 12 min readRead it on haqq.ai
- StudiesGovernance by Construction: AI Guardrails You Can't BypassSix popular LLM guardrails were bypassed at 12–100% rates. Governance by construction builds the unsafe action out of the agent - nothing left to evade.May 2026 · 12 min readRead it on haqq.ai
- StudiesAI Hallucinations in Law: 1,313 Court Cases and Counting1,313 court proceedings, 496 sanctioned attorneys, five enforcement tracks across 106 countries. The global report on AI hallucinations in law.May 2026 · 45 min readRead it on haqq.ai
- StudiesPrompt Injection in Legal AI: 5 Attacks, 5 Blocks, 1.84 msFive adversarial NDAs, five prompt-injection payloads, one zero-dependency scanner - all blocked in under 2 ms. How injection hits legal AI and what stops it.May 2026 · 9 min readRead it on haqq.ai
Workforce and outlook
Role transformation and the forward view of legal work.
Nothing in the library matches that search.