After more than a year in stealth, Bono Network announces Arbiter, a universal document intelligence engine that replaces prompt-wrangling with structured, reliable outputs designed for everyday legal work. Arbiter addresses persistent Legal AI failures which legal professionals often encounter whilst utilising AI tools to assist them in their tasks, including lost-in-the-middle, context rot, hallucinations and usability gap, through a segmented, multi-pass architecture that surfaces risks, redlines, and recommendations.
Unlike generic chat interfaces, Arbiter is built to deliver structured reports and JSON that slot directly into counsel, compliance, and CLM workflows, or easily exported into word and pdf documents, cutting review time and improving repeatability across teams.
“Lawyers don’t want to ‘talk to an AI’, they want clear, auditable outputs they can use,” said Ilya Arbabi, Founder & CEO of Bono Network. “Arbiter is engineered to think in segments, verify across sources, and return structured answers, so teams get clarity, enforceability, and risk calls they can rely on.”
Arbiter’s features
Arbiter presents several features which distinguish the product from anything else available on the market:
- Universal Document Intelligence (multi-language, multi-jurisdiction): deep analysis of contracts, policies, and filings with clarity, enforceability, risk, and recommendation scoring, plus section-by-section flags, cross-reference detection, key date tables, and financial summaries.
- Accuracy centric architecture (by design): segmented vector reasoning, semantic anchors, and multi-pass validation to prevent context drift and “needle-in-the-haystack” misses on long documents.
- Deliberation Mode: a slow, deep-thinking option where internal agents (Abaddon, Anaïs, and Arbiter) debate and converge. In practice, this allows Arbiter to review hundreds of sources in one run and return validated, cited outputs for complex research tasks.
- Legal Research Module: performs live, jurisdiction-aware research, builds chronological timelines, tables, decision trees, and entity graphs on request, so teams get analysis and artefacts, not just prose.
- Multi-Document Workflows (private testing): portfolio-level review of matter bundles and playbooks with roll-up risk matrices for partners and GCs.
Multi-jurisdictional research
One of Arbiter’s points of strength is its ability to conduct multi-jurisdictional research without making the mistakes which are common to AI softwares.
Independent benchmarking this month found that all evaluated AI products and even lawyer baselines struggled most on complex, multi-jurisdiction questions, with average scores 14 points lower than single-jurisdiction tasks; 50-state surveys were the weakest category for everyone due to sourcing and citation challenges. Arbiter’s specialist workflows, designed to pull and validate authorities across many jurisdictions, are purpose-built to close this gap.
Early results and real-world impact
In early use, Arbiter identified critical issues missed by multiple reviewers at three firms, including one matter where it averted approximately $400,000 in potential exposure by surfacing a buried risk and recommending targeted edits.
Practitioners evaluating Arbiter have also highlighted its differentiators versus well-known tools in the space, specifically, the depth of the section-by-section analysis, risk triage, and report quality, citing a more comprehensive output than “prompt-only” workflows.
“Arbiter shifts legal AI from conversations to decisions, allowing users to get a report you can file, as opposed to the common chat transcript which ultimately needs to be fact-checked or further researched.
Availability
Arbiter is available today as an application (seat-based) and as an API that returns structured JSON and reports. Private multi-document workflows are in testing with design partners. Interest forms and demo requests are open now at hirearbiter.com and users can also try Arbiter for free on our website https://arbiter.bono.network.
