Pauhu and the Pro-Human AI Declaration
Pauhu agrees with the spirit of the Pro-Human AI Declaration. The text is the Future of Life Institute's, and we do not copy it. Read it at its source.
Status: we have signed. The site checks each signature before listing it, so we will say "endorsed" only when our name appears on the public list.
Levels: in place means you can check it today. Partial means some of it is true. Proposed means our intent, not yet checkable.
1. Keeping humans in charge (partial)
- Prepaid balance does not renew. You top up when you choose, and no use means no charge. Our terms say so; payments are not open yet, so this is proposed until they are.
- Our open tools run on your own device on plans you grant. Partial.
- Self-improvement is done on the backend only, not in the product. Customer memory is not used for it. Proposed. This rule alone does not show we meet the declaration's caution on self-improving systems.
- A clear off-switch for the product. Proposed.
2. Avoiding concentration of power (partial)
- Code and texts in our open repositories are Apache-2.0. In place.
- Our device-side tools pages link to the upstream makers and host no weights. In place.
- A free builder key is offered on our developer site. Offered; we are still fixing its sign-in.
- Not everything we build is open.
3. Protecting the human experience (partial)
- No ads on our sites. In place.
- Labels that say a result is AI-made, and that the AI is not a person or a professional. Proposed.
- No manipulation to keep you using it. We found no such tactic on our pages, but we have no test to show. Proposed.
4. Human agency and liberty (partial)
- Our terms promise each answer carries its official source, or a clear note when a source cannot be verified. Policy in place; not yet tested across every answer.
- Our privacy page lists your right to see, correct and erase your data. We answer within 30 days. In place.
- We do not sell personal data. In place.
- Analytics cookies are set only after you accept. In place.
- Deleting your data from training sets and derived data. Proposed.
5. Responsibility and accountability (partial)
- The seller is named with its business ID and contact routes. In place.
- Our terms say we surface and cite sources and do not guarantee outcomes. In place.
- Decisions made in public issues and pull requests. Proposed.
- A way to find out why an answer was given and who is responsible. Proposed.
Questions and answers
- Did Pauhu sign?
- Yes. Our signature is waiting for the site to check it. We will say endorsed only when our name appears on the public list.
- Does Pauhu AI improve itself?
- Self-improvement is done on the backend only, not in the product. Customer memory is not used for it. This is our rule today. It is not yet described on our public pages (proposed).
- Is my data used to train or improve it?
- Your questions are used to return your answers. We keep only token counts for API and tool calls. Customer memory is not used for self-improvement. Our privacy page does not yet say anything about training on your queries, so a full statement is proposed.
- How do I see the source?
- Our terms promise every answer comes with its official source and date, or says when a source cannot be verified. Bulk data rows carry their official origin and date. Policy in place; not yet tested across every answer.
- How do I switch it off or delete my data?
- Once payments open, the balance will not renew, so stopping will be as simple as not topping up (proposed). To see, correct or erase your data, use the contact route on the privacy page. We answer within 30 days. A one-step off-switch and delete button is proposed.