Creomind Data Charter

1. Consent is architectural

Nothing is collected, processed, or modeled without explicit, informed, revocable consent. Consent is documented, scoped to a defined purpose, and re-confirmed whenever the purpose changes. Privacy is not just a promise we make, it is built into how we structure access to source material, archives, and Cognitive Models. An archive and Cognitive Model are only ever built from material captured with the direct, informed, revocable consent of a living Source; we do not construct a new archive or Cognitive Model from posthumous material alone. After a Source's death, their designated representatives or estate govern access, continuity, and deletion of the existing archive and Cognitive Model, strictly according to the wishes the Source recorded in life.

2. The Source is the owner

Source material, including voice recordings, writing, photographs, journals, and interviews, and the archive built from it, belongs to the Source or their estate. We act as custodians, not owners. Creomind retains its own underlying software, methodology, and technology used to build the Cognitive Model. The Source or their authorized representative may request export, restriction, or deletion of their data at any time.

3. Minimum viable collection

We collect only what is necessary to construct, validate, and maintain the archive and Cognitive Model agreed upon in the engagement. We do not collect "extra" data on the chance it becomes useful. When a piece of data is no longer needed for a defined purpose, it is deleted.

4. No training on outside models

Source material is not used to train, fine-tune, or evaluate any model offered to a third party, any general-purpose foundation model, or any commercial dataset. Each Cognitive Model is built for one family, from one Source's authorized archive material.

5. No sale, no advertising

We do not sell source material or archive data. We do not share it with advertisers, brokers, or analytics platforms, and we do not share it with any party outside the processors strictly necessary to operate our infrastructure. We will not run advertising against an archive or Cognitive Model.

6. Architecture for security and separation

Current practice: Creomind's infrastructure is hosted on Microsoft Azure. Data transmitted between users and Creomind is protected using encrypted connections such as HTTPS/TLS, and applicable data stored within Creomind's Azure infrastructure is protected using encryption at rest provided by the underlying Azure services. Multi-factor authentication is used to protect administrative access to Creomind's Azure environment and relevant administrative accounts. Creomind does not operate its own physical data-center infrastructure, and we do not present Microsoft's controls or certifications as our own.

Design principle: Creomind is being designed around data minimisation, restricted access, separation of private archive data from development and research environments, and lifecycle controls appropriate to the sensitivity of the information. Formal access restriction and logging, stronger isolation between engagements, and documented operational security procedures are intended practice as our infrastructure matures, not claims about today. Creomind has not commissioned an independent penetration test of its application, and we do not present platform-level provider testing as equivalent.

7. Right to be forgotten

On a verified deletion request from the Source or authorized representative, Creomind deletes or schedules deletion of the applicable personal data from active systems, in accordance with its technical and legal retention requirements. This includes the source material and the derived artifacts used to operate the Cognitive Model, such as model weights, prompts, and embeddings held in active systems. Residual copies may persist temporarily in backups where technically necessary and are removed according to the applicable backup lifecycle. We will confirm in writing when the deletion we can perform has been completed.

8. Transparency on what the Cognitive Model knows

Families receive a plain-language record of the categories of data used to construct the Cognitive Model, the boundaries we configured, the topics it is designed to engage with, and the topics it is designed to refuse. There are no hidden capabilities.

9. Dignity over fidelity

Where a tradeoff exists between technical realism and the dignity, safety, or stated wishes of the Source, dignity wins. We will reduce capability, refuse a feature, or end an engagement before we will compromise on this.

10. No impersonation in the world

Cognitive Models are private instruments. They are not deployed into public channels, social media, customer service, political messaging, or any context where their generative output could be mistaken for the living or speaking-on-behalf-of the Source.

11. Independent review

Creomind does not currently undergo independent external security, privacy, or ethics review. As Creomind develops, we intend to introduce appropriate independent review of selected security, privacy, and governance practices. Findings that affect families will be communicated to them directly.

12. This Charter is binding

This Charter is incorporated into our Terms of Service and our Privacy Policy. If any of our conduct ever falls short of it, the Charter is the standard we will be held to, by you, and by us.

Questions, concerns, or requests: info@creomind.ai.