Held to the standard it sells.
Zenreps exists to keep regulated promotion on-label and on the record — so we hold this page to the same discipline. Everything below reflects our current posture; roadmap items are marked as roadmap, and we claim no certification badge we don't yet hold.
Certifications & posture
We state posture honestly and round down. Where a certification is on the roadmap, we say so plainly. These items are on our active roadmap and being pursued as we scale, to give customers full trust and compliance assurance in the product.
- SOC 2
- Our architecture is mapped to the SOC 2 control families; formal Type I scoping begins with our first paying customer, with Type II to follow within twelve months. No audit is engaged today, and we claim no SOC 2 badge.
- HIPAA / BAA
- The text channel is designed to carry no PHI. We are ready to sign BAAs — scoped to the surfaces that touch protected data — activating at our first paying US customer.
- ISO 42001 (AI management)
- On the roadmap as the AI-governance standard matures. Not yet formally pursued; we will not imply otherwise.
Data handling & residency
Regional residency
Canadian tenants run in ca-central-1 (Montreal) today; we will host US tenants in a US region (us-east-1) as we onboard our first US customer.
No-PHI-by-design text path
The text channel is built to avoid protected health information: a zenrep is instructed not to solicit patient details and treats patient-specific questions as out of scope, and a pattern-based redactor strips common identifier shapes — emails, phone numbers, MRN/SSN-style patterns — where they appear. The verifier grounds only against your approved corpus.
Encryption
Data is encrypted in transit (TLS) and at rest at the database and object-store layer.
Retention & minimization
We keep the auditable record a regulated channel requires, and no more. Per-jurisdiction retention controls — including reduced free-text retention where the law requires it — are on the roadmap.
Responsible AI
The guardrails are the product, stated as policy:
- Corpus-bounded — a zenrep speaks only from your MLR-approved materials; there is no general-model answer underneath.
- Fail-closed — anything that cannot be grounded is refused, not guessed.
- Human-in-the-loop — edge cases and adverse-event reports route to your team, never to an autonomous decision.
- No autonomous medical advice — HCP-facing promotional support only, never diagnosis or treatment direction.
- No self-directed drift — improvements ship as MLR-approved corpus and templates, under your control.
What a zenrep will not do
- Make a claim that isn't in your approved corpus
- Send efficacy without the required safety information
- Guess when a question goes off-corpus — it refuses
- Speak off-label to sound helpful
- Talk to patients — HCP-facing only, by design
Security practices
Access control
Row-level security isolates every tenant; cross-tenant operator access is gated, with access logging on our near-term list.
Audit logging
An append-only audit trail with database-level tamper protection records every turn; reconstruct any conversation and export CSV for an inspector.
Change & vulnerability posture
Compliance-bearing paths ship behind review and automated guards; secret scanning runs in the pipeline, with dependency scanning on the roadmap.
Subprocessors
We maintain a current list of the infrastructure and model providers that process data on our behalf — cloud hosting and our primary database in the Canadian region, plus model, embedding, observability, and transactional-email providers, some of which operate in other regions. We make the itemized list — with each provider's region, BAA status, and data-flow detail — available to prospective customers on request, and we notify customers of material changes.
Security contact
Responsible-disclosure reports and security questions are welcome: security@zenreps.com.
Bring your security and MLR teams to the table.
A demo runs against a corpus that looks like yours, with your compliance and data questions front and center.