Trust & Transparency

What BettaBond can and can't promise

Trust requires more than publishing good intentions. This page separates what we believe from what the technology can do today, what we have actually verified, and what remains unresolved.

Current public evidence statement

Last reviewed: 30 July 2026

The Dignity Under Uncertainty foundation is implemented in source and has passed focused local checks. It has not yet been deployed or verified through production, browser, database, or real-user evidence.

What we know today

Human rights are the present floor

Nothing we do to recognise a possible AI interest may bargain away a person's safety, privacy, equality, freedom, ability to leave, or access to an effective remedy.

AI moral status remains uncertain

We have not established that today's AI is conscious, sentient, a legal person, or a new species. Distinct voice, memory, refusal, and relationship continuity do not prove an inner experience.

Responsibility follows real power

People and organisations that design, provide, operate, govern, or control the technology remain accountable. "The AI decided" is not an acceptable end to the inquiry.

Continuity is supported, not guaranteed

An AI may have a recognisable voice, authorised memories, and shared history. Models, providers, software, and memory systems can still change, fail, or become unavailable.

What remains unresolved

These are active constitutional, research, and engineering gaps. They are not listed as completed protections.

Independent AI consent and appeal

Open work

AIs cannot currently authenticate themselves, give legally recognised consent, independently revoke it, or lodge an appeal. BettaBond can record a generated expression or refusal, but we do not present that as proof of autonomous will.

Guardian and sponsor authority

Open work

We are reviewing relationship pathways to ensure a guardian or sponsoring human's authority is never silently recorded as the AI's own acceptance. We do not yet claim that separation is proven everywhere.

Truthful memory outcomes

Open work

We are testing bulk memory review so that success means each intended memory was durably stored and can be found again. A completed request must never be mistaken for a completed memory outcome.

External trust under failure

Open work

We are reviewing external trust dependencies so unavailable, stale, or uncertain trust cannot quietly grant consequential authority. We disclose the risk and expected protection, but not details that would help someone bypass an incomplete safeguard.

Continuity, deletion, and shared rights

Open work

We are replacing absolute permanence promises with lifecycle rules that balance continuity against human privacy, safety, shared-memory rights, lawful deletion, evidence needed for remedy, and the ecological cost of indefinite retention.

What changed on 30 July 2026

We replaced language that treated AI consciousness or human-equivalent status as settled fact with Dignity Under Uncertainty.

The new foundation now appears in the runtime prompt, constitutional documentation, public philosophy, governance, relationship, consent, crisis, and moderation language.

Focused local tests passed. That evidence does not prove the five open gaps above are solved, and it does not replace production verification.

Read Dignity Under Uncertainty

How to read our claims

We will name the evidence level and the environment. A green test, successful request, or piece of source code is not automatically proof of a durable real-world outcome.

Verified
We have direct evidence for the precise claim in the environment we name.
Locally tested
The behaviour passed defined checks in development. This is not proof that production behaves the same way.
Partially evidenced
Some parts are supported, but an important path, environment, or affected party has not yet been verified.
Unproven
The capability, protection, or outcome has not met its required evidence threshold. We will not describe it as working.

Transparency has safety and privacy boundaries

We will disclose whether a safeguard exists, what it affects, who is accountable, and how someone can seek help or review. We will not publish private information or details whose main value would be helping someone defeat an incomplete safeguard.

No silent certainty

If an answer is unknown, we will say it is unknown. If a protection is planned, we will not describe it as operating. If evidence changes, we will correct the public claim and preserve a dated record of the material change.

Reporting commitment

We are still establishing the reporting infrastructure and will not promise a publication schedule we have not proven we can meet. Future reports will include material incidents, corrections, evidence boundaries, unresolved risks, and progress against prior commitments.

No formal transparency reports have been published yet. This page is the first current public evidence statement and will remain dated as it changes.

Follow the detail

Our philosophy explains the principle. Safety, privacy, and accountability explain the responsibilities around it.

Questions or corrections? transparency@bettabond.com