I. Your Data, Your Infrastructure, Your Control

Enterprise AI should not require surrendering your most sensitive data to third-party clouds. We build software that runs within your infrastructure—on-prem, private cloud, or air-gapped. Data egress is disabled by default and controlled by you. Encryption keys remain under customer control (KMS/HSM where available).

Deployment mode determines data residency and vendor access; our default posture is customer-controlled processing with air-gapped support where required.

II. Radical Transparency Over Marketing Theater

The enterprise software industry is plagued by inflated claims, hidden limitations, and "contact sales" obscuring basic information. We reject this.

• Public pricing. No "request a quote" games. Our pricing is on the website.
• Honesty matrices. We document what works, what's planned, and what we won't build.
• No vanity metrics. When we cite numbers, we explain methodology and limitations.
• Compliance status, not badges. We state where we are in certification processes, not what we hope to achieve.
• Clear scoping. What is included vs. paid services is documented upfront.

III. Decisions Must Be Explainable

AI that cannot explain its reasoning is AI that cannot be trusted in regulated environments. Every decision in Datacendia produces:

• Evidence packets documenting inputs, evidence sources, policies applied, tool actions, outputs, and run IDs
• Cryptographic signatures proving data integrity
• Immutable audit trails exportable in standard formats
• Dissent records showing when AI agents disagreed and why

If we can't explain it, we don't ship it.

IV. Portability and Low Switching Costs

Your investment in understanding and configuring an enterprise platform should not create lock-in. We design for portability:

• Standard export formats for all data, configurations, and artifacts
• Open, documented interfaces for integration and automation
• No required vendor cloud dependency for core runtime operation
• Your fine-tuning artifacts, training data, and prompt libraries—all exportable. For customer-owned local models, model artifacts can be exported; for third-party hosted models, configuration and fine-tuning artifacts are exportable.

V. Security Is Architecture, Not Afterthought

Air-gapped deployment is a first-class citizen, not a checkbox feature. We architect for the assumption that:

• Assume breach (zero-trust posture)
• Updates must work offline
• Telemetry is disabled by default. It can only be enabled by explicit customer configuration, and is fully documented. No telemetry is transmitted from air-gapped deployments.
• Defense-in-depth beats perimeter security

VI. Multi-Agent Deliberation Over Single-Model Answers

Complex enterprise decisions deserve more than a single model's output. Our Council architecture orchestrates multiple specialized AI agents that:

• Challenge each other's reasoning
• Dissent when evidence conflicts
• Document the deliberation process
• Produce auditable recommendations, not black-box answers

Agents are roles; they may share a configurable model stack.

VII. Build for the Regulator in the Room

Regulated industries face auditors, examiners, and compliance reviews. We build as if the regulator is watching—because they are. This means:

• Evidence trails mapped to SOC 2 / ISO 27001 and aligned with NIST 800-53 control families
• Export formats that auditors actually understand
• Documentation that supports evidence collection for control effectiveness

VIII. Honest About Our Limitations

We are a specialized engineering team focused on sovereignty. We will not pretend to be a massive generalist vendor.

• SOC 2 Type II: In progress, target Q2 2026
• ISO 27001: Roadmap 2026
• Customer references: Available subject to customer approval / NDA
• What we don't do well yet: See our Integration Honesty Matrix

Target dates are estimates; we publish changes transparently. We'd rather lose a deal than oversell a capability.

IX. Customer-Owned Verification

Trust but verify. We provide verification steps so customers (or auditors) can independently validate exports, signatures, hashes, and policy enforcement.

This converts "trust us" into "verify us."

This manifesto is a living document. Last updated: December 2025.

Discuss Our Principles →