Moderation Policy

Last updated: September 2, 2026

APIbase is a tool for developers and software agents. It is not a consumer service.

Every call to an action-class tool — one that sends a message, places a call, or posts content on your behalf (currently: Telegram, Twilio, Telnyx, Resend) — is scanned automatically against a fixed prohibited-content ruleset before it reaches the provider. Every other tool (search, data lookups, and everything else) is checked only against one absolute rule: child sexual abuse material (CSAM). See Terms of Service §5 for the full list of prohibited categories.

What happens when a call is blocked

• The provider is never called — the block happens before any upstream request.
• If the tool call was free, the request is simply rejected. Nothing is charged, nothing is stored.
• If the tool call was paid, payment for that call is still charged (escrow settles at rejection time, the same as a successful call) — this is what makes an appeal meaningful: there is a real charge to contest, not a hypothetical one.
• A paid block's response includes an appeal_id, a link to the appeal page, and a link back to this page.

What we keep

For a blocked paid call only — never for a free one, and never for anything that wasn't blocked — we retain:

• The full value of the specific field that triggered the rule (capped at 4KB; a longer field is truncated and marked as such), and where in it the match occurred.
• Exception, with no exception to the exception: child sexual abuse material (CSAM) is never retained, in any form, at any length. A CSAM block keeps only the rule id, category, and appeal id — never the content itself. Review of a CSAM block is done from those identifiers alone.
• This is retained until 30 days after your appeal is resolved, if you appeal within 14 days of the block. If you never appeal, it is deleted automatically 14 days after the block.
• After that, only the skeleton — rule id, category, appeal id, and the resolution verdict — is kept, indefinitely, for audit purposes.

We keep the full field, not a short excerpt, on purpose: a few words pulled out of context (e.g. "crisis support") can look identical to a genuine violation and identical to a false positive. Reviewing the full field is what lets a human tell the difference.

Access to retained content is limited to the appeal page itself (you already sent it — this is what lets you and a human reviewer see exactly what was flagged) and an operator-only review tool. It is never written to application logs or sent to any messaging/alerting channel.

Our failures are different. If your payment succeeds but the upstream provider fails to deliver, that is not a moderation block — a refund is owed without any appeal. Don't confuse the two: a moderation block names its appeal ID; a provider failure does not.

Appeals

• A paid block's response gives you an appeal_url — open it to see the block reason and, if applicable, the flagged content, and to submit your side of the story.
• You may appeal within 7 days of the block.
• A human reviews every submitted appeal within 72 hours and marks it upheld (the block stands) or overturned (the block was wrong).
• If the block was our error, we refund or credit the full charge — that is owed, not discretionary. A balance credit is applied promptly and is the default; an on-chain refund is processed manually and may take up to 14 days after the appeal is upheld.

Repeat blocks

Repeated blocked calls within a short period suspend further service, regardless of payment status, independent of any individual appeal's outcome.