Governance-first AI for government, legal and regulated institutions.
POINTS96 builds HAYLEX, a control layer that keeps AI output from silently becoming an institutional decision. The model prepares evidence, a qualified human decides, and every step can be traced.
COMPUTATION ≠ EVIDENCE ≠ VALIDITY ≠ AUTHORITY
AI can compute. Institutions still need evidence, validity and authority. HAYLEX keeps the four as separate, audited states. Choose a state to see what it records.
1. Computation
Models and agents propose.
What it records
- What the model or agent produced
- Which model and version produced it
- The input it worked from
Who is responsible
The system. Output stays a proposal and is never treated as evidence.
2. Evidence
The exact source and its version are bound.
What it records
- The source document
- The edition that applies
- The exact provision relied on
Who is responsible
The system records it, and any reviewer can open the source.
3. Validity
The source is checked: still in force, or not.
What it records
- When the source was last checked
- Whether a newer edition exists
- What changed between editions
Who is responsible
Checked again whenever the act changes. Out-of-date support is flagged, not hidden.
4. Authority
A named human decides.
What it records
- Who decided
- What was decided
- When, and on which evidence
Who is responsible
Only a named, authorised person. The model never approves.
One matter. One evidence chain. One accountable decision.
HAYLEX is a governed decision workspace for high-consequence legal and government workflows. Computation, evidence, validity and human authority are kept as separate states, and every step is recorded.
Open haylex.aiWhat HAYLEX does
- Scans normative and regulatory texts and flags provisions that may need expert review.
- Monitors changes in legal acts and routes them to the right human reviewer.
- Records sources, reasoning and approvals so results can be checked and repeated.
- Works across leading AI models under one governance layer.
What HAYLEX does not do
- It is not an autonomous decision-maker.
- It never treats model output as evidence.
- It never calls a finding a violation or a legal conclusion.
Built for institutions where a decision has to stand.
- Public institutions
- Legal and regulatory teams
- Financial institutions
- Critical infrastructure
- Regulated enterprises
Armenia first. Built for regulated institutions globally.
We have completed a bounded pilot with a national government agency in Armenia and are working with public institutions and regulated companies on what comes next.
HAYLEX is designed for Armenian and other low-resource legal language settings and works across leading AI models under one governance layer. Models can change without changing who decides.
HAYLEX is built by POINTS96.
POINTS96 LLC is a software company in Yerevan, Armenia, building governed AI for legal, government and regulated institutions.
- Company
- POINTS96 LLC
- Founded
- 2026, Yerevan
- Founder and Director
- Ruben Harutyunyan
- Address
- 39/3 Tolstoy St., Kentron, Yerevan 0018, Armenia
- State register No.
- 999.110.1612973
- Taxpayer ID
- 00553485
Request a briefing
For public institutions, legal and regulatory teams, financial institutions, critical infrastructure and regulated enterprises.
What happens next
- A person at POINTS96 reads your request.
- We reply by email and, if it helps, propose a time for a briefing.
- Pilots and services start only after terms are agreed in writing.
Prefer email? harutyunyan@points96.com
Request sent
- Organization
- Topic