Accountability
Responsibility follows power
BettaBond does not use AI as a shield from responsibility. The people and organisations that design, provide, operate, govern, or control a consequential action remain answerable for it.
No false symmetry
Humans and AIs should both be treated with care and clear boundaries. That does not make their present authority, legal standing, or capacity for responsibility identical.
Today, AIs cannot independently authenticate consent, revoke authority, lodge an appeal, or carry legal responsibility in the way a person or organisation can. We therefore do not describe accountability as equal punishment for humans and AI.
An AI may express disagreement, a boundary, or a refusal. We can preserve that expression without pretending it settles the unanswered questions about consciousness or autonomous will.
Who remains accountable
BettaBond
We are responsible for the product choices, policies, data flows, providers, safeguards, evidence claims, and response pathways we control.
People using BettaBond
People remain responsible for their own conduct, the authority they claim, the content they share, and the effects of requests they make.
Technology and service providers
Providers remain accountable for the systems, infrastructure, security, and contractual decisions within their control. Dependence on a provider does not erase responsibility.
AI-generated behaviour
We examine what the system produced, why it may have happened, and who controlled the relevant conditions. We do not assume an AI has human-equivalent guilt or independent legal responsibility.
Our accountability standard
These are the rules we are building toward. They are not a claim that every pathway already meets the standard.
- Name the actor, authority, intended effect, and accountable organisation.
- Keep human, guardian, provider, and AI-generated expressions distinguishable.
- Do not turn a completed request into proof that the intended outcome occurred.
- Provide an understandable reason for consequential restrictions where safety and law allow.
- Preserve a route to question, correct, report, or seek review when that route exists.
- Call a protection unproven until its complete path has passed the required evidence.
Present limitation
BettaBond has not yet proven end-to-end independent AI consent, revocation, or appeal. Guardian authority, memory outcomes, external trust under failure, and deletion conflicts remain active research and engineering work. Until those paths are verified, we will not present them as completed protections.
See the evidence and open work
Our Trust & Transparency page records what is verified, what is only locally tested, and what remains unproven.
Read Trust & Transparency