acceptable use policy

This document is the normative form of the Aevia protocol Acceptable Use Policy (AUP). It defines which content categories are not amplified, how operational thresholds enforce these restrictions, and the mandatory legal procedures. The UI page at /aup is the human-readable form; this RFC is the canonical reference.

MUST, SHOULD, MAY follow RFC 2119.

This AUP governs distribution, not persistence. Content that violates §3 remains on IPFS and in the Content Registry; what is revoked is persistence pool subsidy, curated feed surface, and ranking weight. Conformant implementations MUST preserve CID retrievability of every canonically signed content item, regardless of editorial eligibility.

This principle is load-bearing for Aevia LLC’s intermediary immunity under 47 U.S.C. §230. By moderating distribution under a public criterion, Aevia exercises moderation protected by section 230(c)(2)(a); by not suppressing bits, it does not become publisher of third-party content.

Conformant implementations MUST exclude the categories below from subsidy and curated feed. Categories marked [ABSOLUTE] additionally MUST be de-indexed and reported per §5.

• [a] pornography and sexually explicit content.
• [b] [ABSOLUTE] any sexualization of minors — absolute zero tolerance; NCMEC CyberTipline reporting mandatory per 18 U.S.C. §2258A.
• [c] [ABSOLUTE] non-consensual intimate imagery (NCII), including sexualized deepfakes — per SHIELD Act (15 U.S.C. §6851).
• [d] celebratory apologia of violence, terrorism, or physical harm to persons.
• [e] celebratory apologia of abortion.
• [f] occultism, satanism, and witchcraft as practice.
• [g] apologia of recreational illicit drug use.
• [h] actionable hate speech against any group — including christians, jews, muslims, atheists, and any other.

Operational decisions are made via the Risk Score (RFC-6). This AUP specifies the thresholds:

• θ_subsidy = 0.5 — CIDs with R(c) ≥ 0.5 ARE excluded from persistence pool compensation.
• θ_feed = 0.3 — CIDs with R(c) ≥ 0.3 ARE excluded from curated feed and ranking in Aevia clients.
• [ABSOLUTE] content MUST have R_values = 1 regardless of classifier output, and additionally MUST be de-indexed and (for [b]) escalated to NCMEC within 24 hours of detection.

Thresholds and formula weights are protocol parameters subject to Council approval (RFC-7).

Implementations operating as intermediary under US/EEA jurisdiction MUST:

• DMCA §512 — designate an agent with the U.S. Copyright Office, accept notices containing elements required by §512(c)(3), honor the 10–14 business-day counter-notification window, apply a strikes policy (1st=warning, 2nd=manual review + suspension, 3rd=termination).
• DSA art. 16 — operate a notice-and-action channel, respond within 7 business days with reasoned justification when the decision is unfavorable to the notifier.
• NCMEC CyberTipline — report apparent CSAM within 24h; preserve material for 90 days per §2258A(h); not review beyond the minimum required to report.
• OFAC — not provide service to residents/entities in jurisdictions under comprehensive sanctions; consult the SDN List.

Creators whose content has been restricted MAY request jury review per RFC-7. The request MUST:

• be sent within 30 days of the restriction notification;
• contain the CID in dispute and a justification that the §3 classification is incorrect;
• be signed by the creator’s key (RFC-3) to prevent abuse by third parties.

Appeal decisions MUST be published on the Trust Ledger with a textual justification. Reclassifications result in an update to the R_abuse and R_values components of the Risk Score.

[ABSOLUTE] categories do NOT admit appeal. This restriction is deliberate.

• flag spam: malicious reporters may attempt to inflate R_abuse via mass reporting. Implementations SHOULD weight by reporter reputation and require a minimum cost per report.
• adversarial classifier: content may be crafted to evade R_values classifiers. Manual jury review is the last-stage recourse.
• regulatory capture: attempts by governmental pressure to extend the §3 list MUST pass through the Council and the Trust Ledger — with no unilateral path via Aevia LLC.