Trust And Assurance

A trust page for buyers who want more than cloud-flavored reassurance.

ChtSafe is shaped by government IT-security-trained practitioners and teams with KRITIS, ISO 27001 audit, and ISO/EN 9001 environment experience. We follow BSI IT-Grundschutz thinking, build with an assume-breach mindset, and keep the trust boundary explicit from hosted rollout to fully air-gapped deployment.

info@chtsafe.com Disclosure policy Privacy policy

Reviewed April 2026

Public trust snapshot

This page is intentionally high-level. It gives buyers the assurance language they expect without turning our public website into an attack blueprint.

BSI IT-Grundschutz aligned Operational posture informed by German public-sector security practice.

Assume breach by default We design for the moment a system gets hit, not for the happy-path brochure.

Layered control walls Critical boundaries are wrapped in overlapping 2-3 deep controls.

Optional session encryption Users can add an additional encryption layer on top of platform controls.

ShinrAI belongs here too: protection starts before third-party inference, not only at storage or transport layers.

Overview

Buyer-friendly trust language, backed by real security engineering habits.

The usual trust-center theater is easy to copy. The harder part is building systems around compromise containment, layered controls, and boundaries that can actually move.

Assurance context

Built by people who have seen real audits and real critical environments.

Government IT-security specialist training informs how we think about control layering and documentation.
Engineering and administrative team members have worked on KRITIS environments and audit-driven delivery.
ISO 27001 audit familiarity shapes our review posture even where we do not market unsupported badge claims.
Process discipline from ISO/EN 9001-style environments carries into change management and operational hygiene.

What we want buyers to understand

Plain language

ChtSafe is not just another shared wrapper on top of a public-cloud LLM tenancy.
ShinrAI changes the request path before provider inference begins.
Hosted, own-instance, dedicated-hardware, and air-gapped delivery are part of the same product story.
Apps, workflows, Team Members, Sims, media, and API all sit behind the same trust model.

What we do not do

No badge theater

We do not replace security review with a cloud logo and a vague checklist screenshot.
We do not publish unsupported certification claims just because competitors do.
We do not expose exact defensive blueprints publicly when procurement review is the proper path.
We do not assume one compromised control should mean full compromise of a protected segment.

Controls

Public control summaries

Detailed implementation is reviewed privately. Publicly, we summarize the control posture that matters to security and procurement teams.

Infrastructure security

Updated April 2026

Live patching is used where the platform supports it; elsewhere we favor staged near-zero-downtime security rollout.
Hardened baselines, access restrictions, and rapid patch handling are treated as default operational hygiene.
Security controls are layered across network, host, runtime, and service boundaries rather than trusted as single gates.
Hosted, dedicated, and on-prem modes do not collapse into one generic infrastructure story.

Containment and segmentation

Defense in depth

We prefer 2-3 wrapped control layers around sensitive segments instead of one decisive wall.
A single compromised component should not automatically imply full segment or tenant compromise.
Operational boundaries are compartmentalized so the blast radius is intentionally smaller than the full platform.
Deployment choices exist because different workloads deserve different trust boundaries.

Cryptographic data protection

Layered encryption

Database, filesystem, and backup protections stack together instead of being treated as interchangeable checkboxes.
Memory handling includes scrambling and aggressive wipe semantics for sensitive data where applicable.
Encrypted retention matters because attackers and operators should not get easy plain-text leftovers by default.
Users can add another layer via session encryption inside the portal when policy requires stronger separation.

Operational discipline

Review-ready

Security review material is prepared around architecture, boundaries, and operating model instead of vague marketing language.
Subprocessor and dependency detail is shared in the context of the actual delivery model under review.
Least-privilege thinking applies to automation, administrative access, and service-to-service credentials.
We continuously assume attackers are capable and adaptive, so controls are built to degrade gracefully under pressure.

Assume Breach

Default assumption: sooner or later, something gets compromised.

That assumption changes engineering decisions. We do not design only for the clean demo path. We design as if an attacker eventually gains more access than anyone wants, which means data still needs protection in memory, at rest, in backups, and between service boundaries.

Protect the data, not just the perimeter: storage, runtime, backup, and routing layers all matter.
Compartmentalize failure: one broken component should not equal total trust collapse.
Use ShinrAI early: the request path itself should reduce exposure before provider inference.
Let users add another layer: session encryption exists because some threat models need more than platform defaults.

Control area	Public stance	Why it matters
Patch posture	Live patching where supported, otherwise staged security rollout with minimal downtime bias	Critical fixes should not wait behind avoidable maintenance inertia.
Wrapped boundaries	Critical segments are protected by overlapping controls, usually two or three deep	A bypassed control should not mean the wall is gone.
Assume-breach design	Systems are designed under the expectation that attackers eventually gain deep access	Data must remain hard to exploit even in ugly scenarios.
Memory handling	Sensitive material is scrambled in RAM and aggressively wiped after load where applicable	In-use exposure matters, not just disks and backups.
Encryption layering	Database, filesystem, backup, and runtime protections stack together	One missed layer should not expose the whole environment.
User-controlled extra layer	Optional session encryption can add another barrier on top of platform controls	High-risk users and regulated workloads can tighten separation further.
Deployment boundary	Hosted, own-instance, dedicated-hardware, and air-gapped modes are available	Different risk profiles need different trust boundaries.

Resources

Where to go next during review

The trust page is the summary. The security page, ShinrAI explanation, privacy language, and deployment intake routes go deeper.

Security proof and request boundaries Review the hosted, own-instance, dedicated-hardware, and air-gapped trust boundary explanation. How ShinrAI changes the path See the protection layer that acts before third-party inference starts. Privacy and retention language Read the public policy view that matches the architectural boundary story. Start procurement and security intake Request deployment-specific review material, architecture answers, and operating-model scoping.

FAQ

Questions trust and procurement teams actually ask

Is this page a certification statement?

No. This page is a public assurance summary. Where formal evidence, deployment-specific questionnaires, or audit-facing detail is needed, we handle that during the actual review process instead of pretending the website replaces it.

Do you really design with compromise as the default assumption?

Yes. That is why we talk so much about segmentation, layered controls, encryption across multiple storage levels, memory handling, wipe semantics, and optional session encryption. The point is to make compromise harder to turn into full data exposure.

Is ChtSafe just another prettier wrapper on top of Azure OpenAI or similar public-cloud AI?

No. The whole product is designed around ShinrAI before inference, explicit deployment boundaries, encrypted retention, and the ability to move into dedicated or air-gapped environments. That is a different trust story from a single shared wrapper layer in front of a public-cloud LLM tenancy.

Can customers add their own extra encryption layer?

Yes. Session encryption is available in the portal for customers who want an additional user-controlled layer on top of platform protections. That option exists because some threat models need more than the default hosted control stack.

Updates

Recent review notes

A lightweight public log of what this trust page is currently emphasizing.

2026-04-08

Public trust center established

The buyer-facing trust page now centralizes assurance posture, control summaries, review paths, and assume-breach framing.

2026-04

Layered-control posture revalidated

Wrapped boundary controls, live-patching posture, and deployment-specific trust segmentation were re-reviewed for current public messaging.