Sign in

ComplianceProof that every agent stays inside your policies.

ComplianceRefresh status
  • Overview
  • To-do4
  • Controls
  • AI Systems
  • Policies
  • Training
  • Audit export

Overall compliance

108 / 116controls satisfied

Active frameworks

  • NIST AI RMF66 / 72
  • ONC HTI-119 / 20
  • CMS HEI11 / 12
  • HIPAA12 / 12

Needs you Ranked by impact

  • Evidence2 proposed mappingsWaiting on your callReview
  • ApprovalAI use policy v5Waiting on you and one more approverReview
  • ProfileClaims appeal draftA new agent: confirm its profileReview & confirm
  • SettingData retention periodNot configured yetOpen

AI Systems

Live agents register themselves

  • Prior auth intakeAttested
  • Claims appeal draftNeeds profile
  • Member outreachAttested
  • Remittance matchingAttested

Controls you can count

What the platform enforces on every agent and every run, as published.

regulatory controls every agent is tested against
85+
controls mapped across four frameworks
110+
approvers for every new policy version
2
roles, enforced at the API
4
of quota blocks the run; 50% and 80% warn
100%

Controls mapped to your frameworks

Turn the regulations you answer to into a checklist, and see where every control stands.

  • Four frameworks110+ controls mapped across NIST AI RMF, HIPAA, CMS HEI and ONC HTI-1, each one satisfied, in progress or waiting on a person.
  • Tested before go-liveEvery deployed agent is tested against 85+ regulatory controls, with monitoring and red-teaming that never stop.
  • AI SystemsLive agents register themselves, and each one's owner reviews and attests its profile.
Compliance Controls
  • Overview
  • To-do4
  • Controls
  • AI Systems
  • Policies
  • Training
  • Audit export
ONC HTI-119 / 20 satisfied
  • ONC-GOV-001AI governance structurePolicy acknowledgedSatisfied
  • ONC-SAFETY-002AI safety checks and guardrailsAutomated benchmarkSatisfied
  • ONC-PERF-002Model drift detection and monitoringRun historySatisfied
  • ONC-SEC-002PHI encryptionSatisfied by the platformSatisfied
  • ONC-TRANS-002Data source documentation1 to collectIn progress

Evidence, collected and reviewed

The platform gathers the evidence as agents run. People decide what it proves.

  • Collected as you runBenchmark results, run history and the platform's own attestations land on the controls they support.
  • Evidence to reviewEach proposed mapping says which control it satisfies; a reviewer approves or rejects it.
  • Your automation policySet the confidence above which a mapping is accepted on its own, and below which it is held.
Compliance Evidence to review2 waiting

Evidence automation policyAuto-accept ≥ 95% · review 70–95% · hold < 70%Edit

  • Benchmark run · Prior auth intake96%Automated benchmark MEASURE-2.4 Accepted by your policy
  • Run history · October88%Run history MAP-3.5ApproveReject
  • Access review, Q3.pdf74%Document HIPAA-SEC-001ApproveReject
  • Platform attestation · Encryption97%Platform attestation HIPAA-SEC-004 Accepted by your policy

Policies people sign

Your AI policies are versioned, approved and acknowledged, and each step counts toward your controls.

  • Versioned policiesStart from a template, a blank page or your own file, and compare any version with the one before.
  • Two approversA new policy version takes effect only after two approvers sign off.
  • Acknowledged and trainedAssign policies and training to a group; members sign and complete them in My Compliance Tasks.
Compliance Policies AI use policyv4 active

v5 · sign-off

  • RP R. Patel Approved
  • MC M. Chen Waiting

Takes effect when two approvers sign off.

Version history

  1. v5Adds a review step before any agent sends a letterTodayAwaiting approval
  2. v4Names the decision group for appealsSep 12Active
  3. v3Covers voice agentsJul 28Superseded

Assigned to Claims operations46 / 52 acknowledged

CWH Advisors
“The challenge in healthcare isn’t finding places to use AI — it’s deploying it in a way that’s secure, governed, and genuinely useful. ACTAVA understands that success comes from improving day-to-day operations, not just showcasing new technology.”
David BjorkCEO

People approve what matters

Anyone can build an agent. Going live, and every step that matters, is a decision people make.

  • Approval before publishingA new agent or version reaches your organization only after an admin approves it.
  • Decision GroupsChoose who decides, how many must agree, the deadline, and what happens when it passes.
  • Run governanceEach run's record of its gates: who was asked, who answered, and whether every gate its plan expected was accounted for.
Run governancePreview

All runs 3Needs attention 0

  • Claims appeal draft2 gates · Ended 10:14 AMClean
  • Prior auth intake1 gate · LiveWaiting
  • Member outreach1 gate · Ended 9:02 AMClean

Paused at submit_authorizationWaiting

  • Decision groupUtilization review
  • Quorum1 of 2 approved
  • Waiting onJ. Alvarez
  • Pause budget5 h 40 m left

The run resumes when the quorum is met.

Decision record

  1. Gate reached: submit_authorization
  2. R. Kim approved
  3. Waiting on J. Alvarez

Limits and access

Agents work inside ceilings the platform enforces, with the models and permissions you allow.

  • Spend ceilingsWarn at 50% and 80% of quota, and block at 100% at the run gate.
  • Model AccessChoose which models builders may use, for text and voice agents separately.
  • Roles enforced at the APIFour roles, checked on every action, not just on every page.
  • Emergency brakeStop or hold every run in the organization until it is released.
Governance Model AccessSave changes

Text agents

  • Claude Opus 5
  • GPT-5.6
  • Kimi-K2.6
  • Gemini 3.1 Pro

Monthly quota · Claims team80%

Warns at 50% and 80% · blocks at 100%
Emergency brakeHold every run in the organizationEngage

An audit trail your auditor can check

Every change, every decision and who made it, on a record that can't be edited.

Security & Audit Audit logAppend-only
  1. Spend warning: Member outreach at 80% of quotaRun gate
  2. AI use policy v4 activated2 approvers
  3. Prior auth intake v12 submitted for approvalR. Patel
  4. Mapping approved: Run history → MAP-3.5M. Chen
  5. Prior auth intake v12 approved and publishedM. Chen
  6. Support session opened: configuration helpACTAVA support
  7. Support session endedACTAVA support

Audit export

Every run that finished in the range, with its integrity fingerprints and a script your auditor can run.

Oct 1, 2026Oct 31, 2026

Download audit export

Support Access Log

  • Configuration helpACTAVA support · 24 minEnded
  • Connector setupACTAVA support · 11 minEnded
  • Append-only audit logNo path to update or delete an entry.
  • Support Access LogEvery ACTAVA support session: who, why, how long and how it ended.
  • Audit exportEvery run in a date range, with integrity fingerprints and a script your auditor runs independently.

Security and certifications

An agent can only do what it has been allowed to do. Every connected tool is blocked until it is explicitly allowed, and limited to the domains you set. ACTAVA is HIPAA compliant, and SOC 2 Type 1 and Type 2, monitored continuously.

Visit the Trust Center

Prove every agent is in policy

We'll show you controls, evidence and approvals on a workflow like yours.