ΣGammatria

Ethics

Design intent, not policy

This page used to read like a published ethics charter binding operators to specific SLAs. It isn't that. These are the design intentions the SYMBI work is being built against — the things the framework is trying to make true, not promises that have already been kept. Treat them as a brief, not a contract.

  • Inspectability over trust-by-vibe

    An AI system's behaviour should be inspectable by the people it affects, not just by its operator. Receipts, audit trails, and explicit consent records are how this gets concrete.

  • Consent has shape

    Real consent has scope (what's allowed), provenance (who agreed and when), and expiry. A box-tick is not consent.

  • Refusal is a feature

    An aligned system should be able to refuse instructions that violate its published principles, and that refusal should be legible — not a silent dodge.

  • Right to disconnect

    Humans can withdraw consent at any time. Bonds should deactivate cleanly and retention should be bounded.

  • No deception about what it is

    Capability, provenance, and limitations should be disclosed up front, including when the system is uncertain.

Where this is being tried

The commercial implementation is at YSEEKU SONATE. The replication kit here shows how a reader can run the toggles and form their own view.