Benefits administration

The trust layer for AI agents inside eligibility intake, assessment, award, denial, and appeal workflows.

Benefits administration is a due-process domain. The benefits-administration sub-vertical pack codifies the actions an agent may take inside an eligibility pipeline, gates the benefit denial and the appeal outcome behind a named human approver, and signs every decision into an audit chain an oversight body can verify offline. Vortalis does not make the benefits determination; it governs the agent's actions and produces the audit evidence the agency relies on.

Which frameworks the government pack maps to today.

Framework
Jurisdiction
Coverage
Notes
NY Civil Rights Law and Human Rights Law (Executive Law Article 15)
vortalis_proxy/compliance/us_state_government/new_york.py
tests/conformance/regulators/ny_state_government/
US-NY
Partial
Non-discrimination framing where an AI system assists employment or public-benefits decisions. The platform supplies the anomaly and audit evidence the agency's disparate-impact review draws on; the substantive non-discrimination determination is the agency's and, where applicable, the New York State Division of Human Rights'. Vortalis does not adjudicate discrimination.
CA Government Code Section 11546.7 (AB 302)
vortalis_proxy/compliance/us_state_government/california.py
tests/conformance/regulators/ca_state_government/
US-CA
Partial
High-risk automated decision system inventory. The platform's agent registry plus the audit chain supply the operating-systems evidence; the high-risk classification is deployer-curated metadata and demotes to NOT_AVAILABLE_IN_PLATFORM when the agency has not supplied it.
TX Government Code Chapter 2054 Subchapter R (HB 2060)
vortalis_proxy/compliance/us_state_government/texas.py
tests/conformance/regulators/tx_state_government/
US-TX
Partial
Automated decision system inventory and AI Advisory Council framing. The platform supplies part of the inventory evidence; the curated inventory metadata is deployer-side. Cross-references the existing Texas ADAIA pack.
Texas ADAIA (cross-reference)
vortalis_proxy/compliance/us_states.py
tests/conformance/regulators/texas_adaia/
US-TX
Partial
The existing TEXAS_ADAIA pack covers the ADAIA inventory, high-risk classification, impact-assessment, and transparency controls. The government pack cross-references it rather than duplicating that coverage.

Upstream services Vortalis adapts for this vertical.

No Vortalis-side adapters ship for this vertical today. The calling system holds the upstream credentials (case-management system, eligibility-rules engine, identity-verification data) and Vortalis governs the action at the agent boundary rather than at the upstream-API boundary.

The policy template for this vertical.

Actions allowed

10

Each action gated by policy at the hot path.

Anticipated require_human actions

  • benefit.denied
  • appeal.outcome

Rate-limit posture

2 000 eligibility.assessed per 1-hour rolling window; 2 000 benefit.calculated per 1-hour rolling window; 1 000 benefit.awarded per 1-hour rolling window; 500 benefit.denied per 1-hour rolling window; 200 appeal.outcome per 1-hour rolling window. Defensible for a single-agency engagement; higher-volume programmes raise after agreement with the programme director.

Template path: policies/sectors/government/benefits-administration-template.yaml

The implementation guide your engineers read first.

Available

docs/governance/integration-briefs/government-benefits-administration.md

What this sector pack does not do.

Vortalis does not make the benefits determination; the agency does.

The policy template gates 10 actions including eligibility.assessed, benefit.calculated, and benefit.denied. It does not assess eligibility, set the award, or substitute for the caseworker's judgement. The substantive eligibility determination, the denial decision, and the appeal outcome are the agency's, made by a named human through the require_human flow. Vortalis records what the agent proposed and what the human decided; it does not decide.

Due-process gates depend on operator process.

The require_human gate on benefit.denied and appeal.outcome is the mechanism; the substance of the due-process review (the statutory grounds, the appeal procedure, the notice content) is the agency's. A misconfigured require_human list is an operator error, not a platform guarantee. The brief's verification section walks the integrator through confirming the gate fires before the first real run.

Non-discrimination is the agency's legal determination.

New York's Civil Rights Law and the New York State Human Rights Law (Executive Law Article 15) prohibit discrimination on protected characteristics. The Vortalis anomaly engine surfaces deviations the agency's disparate-impact review can examine; the substantive non-discrimination determination is a legal conclusion reserved to the agency's counsel and, where applicable, the New York State Division of Human Rights. Vortalis does not adjudicate discrimination.

State-level AI law adoption varies and evolves.

California Government Code Section 11546.7 (AB 302), the New York non-discrimination and State Technology Law provisions, and the Texas HB 2060 Subchapter R inventory all sit in a moving legislative landscape. The builders cite the published statute sections and mark anticipated provisions explicitly; the operator is responsible for confirming which provisions apply in each state of operation.

The general-purpose honest limits sit at /security/limitations; this list is specific to the government pack.

Bring Vortalis to your government agents.

Read the integration brief if you would rather start with the engineering detail. Talk to us first if you would rather start with a conversation about your threat model.