DRAFT, pending counsel review. This document is an internal draft prepared on 2026-04-26 by the engineering team. It has NOT been reviewed by external legal counsel. Do not rely on it for legal advice. Effective date is a placeholder pending sign-off. Apostle Pty Ltd makes no representation that this draft satisfies any specific jurisdictional requirement until counsel-reviewed.

Editorial Policy

Last revised: 2026-04-26 · Version: 0.1.0-draft

1. Curatorial principles

PYLON is a curated short-film platform. We publish films we believe deserve to be seen. We do not run an open feed, we do not surface by engagement metrics, and we do not publish work to fill a calendar. PRD §11.7 names the editorial guardrail explicitly: a single coherent editorial voice, signed off by humans, with a hard floor on quality and brand safety.

Six commitments follow from that:

• Quality and craft over quantity. A small library of considered films is the product. We programme for taste, not throughput.
• Films, not artefacts. We present films as films, not as prompt galleries, generation logs, or tool-chain breakdowns.
• Named human curators. Every decision is made by a named curator. No automated rejection. No automated publish.
• Provenance honesty. We require that AI provenance be true and complete (see AI Disclosure Policy at /legal/ai-disclosure-policy), but we do not require it be flaunted in the work itself.
• Brand safety as a hard floor. Films that fail the brand-safety criterion do not pass regardless of craft.
• Filmmaker dignity. Declines are template-stable, reasoned, and do not editorialise on the filmmaker as a person.

2. Selection criteria

We select on the six-criterion scorecard published at /editorial-policy. Each criterion is scored 1–5 by the assigned curator. Maximum 30. The publish threshold is 22 of 30. Films scoring 18–21 escalate to a senior curator. Films below 18 are declined.

The six criteria, in order:

1. Narrative coherence. Does it tell a story and have a point.
2. Visual craft. Composition, colour, motion quality, AI-artefact rate. Visible technical control beats raw resolution.
3. Human creative contribution. Evidence of selection, arrangement, iteration, direction, post-production, and final cut authority. See §3 below for the cross-reference to the AI Disclosure Policy.
4. Brand safety. Ads can run against it. No slurs, no hate, no targeted harassment, no non-consenting real-person likenesses, no content in violation of the AU Online Safety Act 2021 or US COPPA.
5. AI-slop indicators (inverted). We score against the patterns we are curating against: faceless automation, meme remix, compilation farm, misleading-trailer cuts, infinite-zoom B-roll. A high score means none of those patterns appear.
6. Originality. Adds something to the AI-film canon. We are building a library, not a feed; every title needs a reason to be in it.

3. The AI Disclosure rubric (cross-reference)

Criterion 3 of the editorial scorecard is "human creative contribution." It is scored against the same six-element framework set out in §5 of the AI Disclosure Policy at /legal/ai-disclosure-policy:

1. Visible selection.
2. Arrangement.
3. Iteration.
4. Direction.
5. Post-production work.
6. Final cut authority.

A film that scores high on visual craft but cannot evidence any of these elements will not score above 2 on the human-contribution axis, which puts it below the publish threshold. This is intentional: we do not publish work where the human role is reduced to a prompt.

4. Submission review SLA

We commit to an initial response within 5 business days of a complete submission. "Complete" means: the title, logline, synopsis, classification, the playback master, the poster and stills, and the AI provenance attestation are all present.

Incomplete submissions are not in the review queue. We will reach out to request the missing pieces; the SLA clock does not start until the submission is complete.

The 5-business-day SLA is for the initial decision (publish, escalate, decline, request revision). A revision pass that comes back restarts the clock.

5. Rejection categories

Declines fall into one of three categories. Every decline names the category and gives a one- to three-sentence rationale.

5.1 Technical

• Master fails encoding spec (resolution / aspect / frame rate / bit depth / colour primaries below the published thresholds).
• Audio peaks unmastered or out of broadcast safe range.
• Captions absent for any title with dialogue.
• Poster or still-frame assets missing or below the published resolution.

Technical declines are recoverable: a re-submission with the deficiency addressed restarts review.

5.2 Legal

• Clearance gap on identifiable real persons (no consent on file).
• Music sync without documented licence.
• Trademarked elements without licence.
• Classification mismatch (the submitter selected M for material that warrants MA15+).
• Age-gate evasion (content for which we cannot accept the submitter's age attestation).

Legal declines are recoverable on resubmission with the missing clearance documented.

5.3 Editorial

• Below 22 on the rubric in aggregate.
• Below 3 on the brand-safety axis (hard floor).
• In conflict with the curated taste expressed in the platform's current programming.

Editorial declines are not recoverable on the same cut. We welcome future submissions.

6. Repeat-rejection policy

A creator with three declined submissions in a rolling twelve-month window enters a six-month resubmit cooldown. The cooldown is calendar-fixed: it begins on the date of the third decline and ends six months later. Submissions during the cooldown are not reviewed and are not counted.

The cooldown exists to protect both parties: it stops curators from repeated declines on near-identical work, and it gives the creator a defined window in which to develop a stronger submission. It does not otherwise affect the creator's account standing, there is no strike recorded.

A senior curator may, in writing, waive the cooldown where the creator's next submission represents a materially different approach.

7. Curator authority and escalation

Decisions move through three tiers:

• Curator (role = 'curator'), first-pass review. May publish at ≥22, decline at <18. Escalates the 18–21 band.
• Senior curator (role = 'senior_curator'), second-pass review. Has publish / decline authority on the 18–21 escalation band.
• Admin (role = 'admin'), has override authority on any decision but uses it sparingly. An admin override is recorded with the underlying reason in the audit log.

Senior curators and admins are subject to mandatory two-factor authentication on every authenticated session (packages/trpc/src/routers/two-factor.ts enforces this on the relevant tRPC procedures).

Filmmakers may request a senior-curator review of any decline by emailing [email protected] within 30 calendar days of the decline. We commit to a written response within 14 calendar days of the request. The senior curator reviews the original decision against the rubric; the senior curator may sustain the decline, request a revision, or publish.

8. No automated rejection

Every decline is signed by a named curator. There is no automated rejection model. There is no engagement-bait detection scoring films. We will not delegate the editorial floor to a model, and we will not pretend that automated triage filters are themselves editorial decisions.

We do use software to support the workflow, submission dashboards, scorecards, queues, audit logs. None of that software issues a decline; the human does. The audit log captures actor_id for every state transition, and system:* actor ids are coerced to NULL in the schema (see commit 832ac90) precisely so an automated transition cannot impersonate a human signoff.

9. Restoration after revision

Where a film is declined for a recoverable reason (technical, legal, or a marginal editorial miss), the creator may revise and resubmit. On resubmission:

• The submission is re-scored from scratch. We do not carry forward scores from the prior pass.
• The clock restarts at the date the revised submission is complete.
• The earlier decline does not count toward the three-decline rolling window in §6 if the revised submission publishes.
• The earlier decline remains in the audit trail.

This policy is meant to encourage iteration, not punish it.

10. Publication state and withdrawal

A published title remains public until one of the following occurs:

• The licence reverts under the Reversion Policy at /legal/reversion-policy.
• A DMCA takedown is actioned (see DMCA Policy at /legal/dmca).
• The creator withdraws the title under the Filmmaker Distribution Agreement.
• A senior curator or admin withdraws the title under §11 below.

11. Editorial withdrawal

PYLON reserves the right to withdraw any published title where information that was not available at publication has come to light and that information would have prevented publication had it been available. Examples:

• A clearance gap surfaces post-publication.
• A real-person likeness is shown to be non-consensual.
• An AI provenance attestation is shown to have been false.
• The title is shown to be a substantial copy of a prior work.

Editorial withdrawals are recorded in the audit log with the actor and reason. The creator is notified and given an opportunity to respond. Where the underlying cause is recoverable (a clearance later obtained, a corrected provenance attestation), the title may be reinstated.

12. Transparency

We publish a curated programming page explaining the editorial direction at any given moment. We do not publish the names of declined filmmakers or the contents of declined submissions; that information stays inside the editorial pipeline and the audit log.

Contact

• Editorial: [email protected]
• Senior-curator review requests: [email protected] (subject Senior review request, [TITLE])
• General queries: [email protected]

Version history