STATUS: DRAFT for Luke’s review before it goes to Craig/ExactPay. Interim document to accompany the interim Services Agreement; counsel will produce the definitive version later. Written to be scrupulously accurate and deliberately defensive: it states the real protections, is plain about what we do NOT have, and — per Luke 7/20 — discloses everything that CAN happen (not just what we intend) and disclaims liability for it, to the maximum extent the law allows. Do NOT edit this to claim insurance, SOC 2, certifications, or guarantees we do not hold. Items flagged [REVIEW] need Luke on the framing before sending.
Provider: OBB Holdings LLC, dba HalyardCo. Product: the HalyardCo CRM + AI Agent, hosted on Cloudflare.
We want to be straight with you up front. There is what we strive for — the protections and practices below, which are real and in place — and there is what we can guarantee, which, in early-stage software that handles real data and uses AI, is very little. No provider can promise that software is error-free, that data is perfectly secure, or that nothing will ever go wrong. This document describes both sides honestly: the measures we take, and — just as plainly — the things that CAN happen despite them. Where we describe a risk, we are disclosing it, not accepting liability for it; your remedies and our liability are governed by the limitations and the cap in the Services Agreement.
In transit. All traffic is encrypted with HTTPS/TLS end to end. The application is served only over HTTPS, with HSTS enforced so browsers refuse to connect insecurely.
At rest. Application data lives in Cloudflare’s managed database (D1) and object storage (R2), which encrypt data at rest on Cloudflare’s infrastructure. Documents in the e-signature system are stored in isolated object storage.
Access control. The CRM is private and login-gated; there is no public/anonymous access to client data. Accounts are role-scoped (an admin, a staff user, and a partner see only what their role permits). Administrative actions on the e-signature system require a separate privileged credential that is held server-side only and never exposed to the browser.
Application hardening. Security headers are enforced on every response (HSTS, a Content-Security-Policy, clickjacking protection via X-Frame-Options, and MIME-sniffing protection). Personally identifying fields are masked in the interface. Public-facing endpoints are rate-limited. Internal/source files are blocked from being served. Secrets and API keys are kept in a managed secret store, never in the application code, and a pre-deploy check guards against accidental exposure.
Per-client isolation. Your instance’s data is isolated to your instance.
E-signature integrity. Signed documents are sealed with a SHA-256 cryptographic hash so any later alteration is detectable, and every step (view, identity verification, consent, signature) is recorded in a tamper-evident audit trail with a verifiable certificate.
Your data is used solely to provide, secure, and support the service for you. We do not sell, rent, or license it, and we do not share it with third parties except the service providers listed below that are strictly necessary to operate the service. We do not use your data to train or improve any general-purpose or foundation AI model.
The service relies on these providers to function; your data (and content the Agent processes) passes through them, each under its own terms that we do not control. - Cloudflare — hosting, database, object storage, network security. - Anthropic — the AI model provider that powers the AI Agent’s assistance. Content the Agent processes is handled under Anthropic’s terms and is not used to train their models. - Resend — delivery of transactional email (notifications, signature invitations). - Microsoft (Graph / 365) — ONLY if you choose to connect your own Microsoft 365 account, and then only to your own tenant, using access you grant and can revoke. - In-app chat relay (currently Discord). The in-CRM chat feature relays messages through a private, access-restricted operator channel so a member of our team (or the AI Agent) can respond. Messages you send in that chat transit that third-party channel. Please do not send secrets (passwords, full account numbers, SSNs) through the in-app chat. [REVIEW: confirm whether to keep, harden, or replace this transport before signing; if kept, this plain disclosure stands.]
We build carefully, but no system is perfect. In the interest of complete transparency, here is what can happen even with the measures above in place. We disclose each plainly and, to the maximum extent permitted by law, disclaim liability for it; your remedies and our liability for any of these are governed exclusively by the limitations of liability and the aggregate cap in the Services Agreement. These are not things we intend — they are things that, in honest terms, are possible.
You can export your data in a common, machine-readable format during the term and for [30/90 — per the agreement’s final terms] days after termination. After that window, data is deleted from active systems in the ordinary course and routine backups clear on their normal cycle.
We believe in being straight about our posture: - We are an independent, early-stage provider. The protections in Section 1 are real and in place. - We carry Errors & Omissions (professional liability) insurance ($1M per-claim / $1M aggregate; biBERK, bound 2026-07-20), and we are adding cyber-liability coverage (in underwriting as of 7/20). We do not currently carry cyber-liability coverage in force. We will not state or imply coverage we do not have. - We are not SOC 2 certified. SOC 2 is a formal third-party audit we have not undergone. - We have not commissioned a formal third-party penetration test.
Our recommendation to any client handling highly sensitive records (e.g. raw bank account numbers, full SSNs, complete payroll runs): keep that data in your system of record for it, and use the CRM for relationship, pipeline, notes, and documents. We are glad to put that scoping in writing.
If we become aware of a security incident affecting your data, we will notify you promptly, investigate, and work with you on remediation. [REVIEW: notification window (e.g. 72 hours) is set in the DPA — keep this consistent with the DPA’s final number.]
Everything in this document — the measures, the subprocessors, the risks in Section 4, and the incident process — is subject to the exclusions of indirect and consequential damages and the aggregate liability cap in the Services Agreement. Nothing here creates any separate or additional liability, guarantee, indemnity, or remedy beyond what that agreement provides. In short: we strive for a great deal, we disclose the risks honestly, and what we can guarantee is limited by, and only by, the Services Agreement.
Draft prepared by Relic for OBB Holdings LLC / HalyardCo, 2026-07-18; comprehensive/defensive rework 2026-07-20 per Luke. Interim — not a substitute for counsel’s definitive version. Not for release to ExactPay until Luke approves.