Terms of service

The rules, written so you can actually read them.

This page summarises the binding Master Services Agreement (MSA) between Hirona and paying tenants, and the Acceptable-Use Policy that applies to every candidate. The full signed MSA governs if the two ever conflict.

Last reviewed: 2026-04-17 · Controller: Hirona Pte. Ltd., Singapore

Definitions

Term
Hirona
Meaning
Hirona Pte. Ltd., a Singapore private company, the operator of the Service.
Term
Service
Meaning
The Hirona platform — AI-native candidate intake, voice interview, and scorecard generation — including the web console, the API, and any documentation we publish.
Term
Tenant
Meaning
The organisation that signs up for a paid or free account to use the Service. Rights and obligations bind every user the tenant admits, not just the signer.
Term
Candidate
Meaning
An individual invited by a tenant to be screened through the Service. Candidates are not parties to the MSA; they are bound only by the Acceptable-Use Policy below.
Term
Customer Data
Meaning
Resumes, transcripts, scorecards, and any other data a tenant ingests into or generates through the Service.
Term
MSA
Meaning
The Master Services Agreement — the binding contract accepted at sign-up. This page is the public summary; the signed PDF governs if they conflict.

Acceptable use

Violation of any clause below is a material breach; we may suspend or terminate access without notice. Candidates are bound by the AUP even though they are not MSA signatories.

No illegal or discriminatory use

Do not use the Service to screen candidates in a manner that unlawfully discriminates on any protected basis, or that is prohibited in the jurisdiction where the screening occurs. You are responsible for compliance with local employment law.

No silent AI use

You must disclose AI-assisted screening to every candidate before the interview begins. Hirona fires the disclosure; do not suppress it via the API or the console.

No reverse engineering

Do not attempt to extract model weights, prompts, or decision logic from the Service. Legitimate auditability is covered by the DPIA + audit log export — not by scraping.

No circumventing rate limits

Do not abuse the API, spin up multiple free tenants to bypass paid quotas, or resell access without a written reseller agreement.

No off-label use

The Service is designed for employment screening. Do not repurpose it for academic admissions, credit decisions, or any other high-stakes domain without a separate written agreement.

Fees + billing

Current plans and unit prices are on the pricing page. Platform fees bill monthly in advance; metered usage (per completed interview) bills monthly in arrears. Invoices are due net-30.

We use Stripe for card + invoice processing. SGT, GST, and VAT are applied where required. Failed payments put the tenant into a 14-day grace window; after that, new interviews are paused until the balance clears. Historic data is always retrievable, even during a pause, so you can satisfy candidate erasure requests.

Your data, our role

For candidate data, the tenant is the controller and Hirona is the processor — the DPA you accepted at sign-up governs. For tenant-admin account data, Hirona is the controller and the privacy notice applies. In both roles, we commit to the following:

  • Residency: Cloud SQL + GCS + KMS pinned to the GCP region the tenant selects at sign-up. Cross-border transfer needs candidate consent and SCCs.
  • Subprocessor notice: 30 days before adding or removing a subprocessor. Subscribe at subprocessors@hirona.ai.
  • Breach notification: within 72 hours of confirmed personal-data breach, matching GDPR Art. 33 timing for all tenants regardless of jurisdiction.
  • Return on exit: structured export of all Customer Data within 30 days of termination; irreversible deletion 60 days after, except the pseudonymised audit records the EU AI Act requires us to retain.

AI-specific commitments

No training on Customer Data

We do not fine-tune, distil, or otherwise use Customer Data to train general-purpose models — ours or any subprocessor's. LLM calls are configured with `store=False` against subprocessors that support it; the subprocessor list is on the Trust center.

Output ownership

Scorecards and transcripts generated from Customer Data are Customer Data — you own them. You grant Hirona a narrow license only to host, process, and analyse them to provide the Service.

Replayability window

Every AI decision is replayable for 7 years against the exact model version + prompt hash used at the time. If the upstream model is deprecated we keep the fingerprint plus our tool-call trace so the decision can be re-explained.

Model routing disclosure

Which provider a specific inference hit is logged and available to tenant admins. We may route between providers for resilience; residency pins override routing.

Human-review SLA

When a candidate exercises their right to human review, we escalate within 48 business hours and deliver a re-scored decision without the original AI output influencing the reviewer.

Warranties + disclaimers

We warrant

  • Uptime targets and credit schedule defined in the MSA, with credits applied when targets are missed.
  • Security controls matching SOC2 Type I; Type II in preparation.
  • Decision audit trail retained 7 years and retrievable on demand.

We disclaim

  • Guarantee that any specific candidate will or will not advance. Hiring is your decision; we provide decision-support.
  • Any implied warranty of merchantability or fitness for a particular purpose beyond the express commitments above.
  • Liability for third-party upstream outages (LLM provider, GCP region), beyond the SLA credits expressly stated.

Limitation of liability

Aggregate liability is capped at 12 months of fees paid, except for breaches of confidentiality, data protection obligations, indemnification duties, or gross negligence / willful misconduct — which are uncapped as required by law. Neither party is liable for indirect, incidental, or consequential damages.

We indemnify tenants against third-party claims that the Service as provided infringes an IP right. Tenants indemnify us against claims arising from their misuse of the Service or violation of the AUP.

Term + termination

How it starts

Click-through acceptance at sign-up creates the MSA. Enterprise tenants may supersede with a signed order form.

How it ends

Either party may terminate for convenience on 30 days written notice (no refund of pre-paid fees), or immediately for uncured material breach after 14 days notice. We may suspend access for AUP violations without prior notice.

Governing law + disputes

Governed by the laws of Singapore, without regard to conflict-of-laws rules. Disputes resolved by arbitration at the Singapore International Arbitration Centre (SIAC) under the SIAC Rules, seat Singapore, language English, one arbitrator. Either party may seek interim injunctive relief from any court of competent jurisdiction.

Consumer-protection rights candidates have under their local law (GDPR, PDPA, CCPA, DCWP enforcement, etc.) are not waived by this clause.

Changes to these terms

Material changes are announced 30 days in advance via email to the tenant billing contact and posted on the Trust center changelog. Continuing to use the Service past the effective date is acceptance. Non-material edits (typos, clarifications, link updates) are logged with a bumped last-reviewed date.

Legal, contract, or negotiation.

Enterprise redlines, order forms, DPA negotiation, or anything where the signed PDF matters more than the summary above.

legal@hirona.ai