Argil moderated content rules vs Sume content_policy_rejected

Argil's moderated-content page restricts finance, crypto, gambling promotion, political topics and misinformation for Argil's own avatars. Sume publishes no topic list of that kind in the sources read here; what it documents is the outcome: a provider can reject generated clip content, and the job then fails with public_reason content_policy_rejected.

Argil rules are quoted from its page, read 2026-10-01; Sume facts are from the OpenAPI schema and Avatar video previews. This is not legal advice.

What does Argil restrict, and for whom?

The page splits three cases: videos with your own avatar have no content restricted; Argil's human avatars (Legacy) are fully restricted; Argil's AI-generated avatars (Atom) skip the entries marked with an asterisk. Those starred entries are finance, illicit promotion, misinformation and political topics. It also says moderation for fictional content is done by third parties and that failed generations are automatically refunded.

What does a rejection look like on Sume?

The public_reason field is a machine-friendly value that SDKs, agents and support tooling can group on; examples in the schema include generation_rejected, image_content_rejected and content_policy_rejected. The multi-scene avatar schema adds that a request of valid length (4-60 seconds in total) can still fail with content_policy_rejected when a provider rejects generated clip content.

How do the two models of moderation compare?

How should I handle a regulated-topic ad?

Check the rejection path before volume. Branch on public_reason, not on message text; surface it to the editor; and do not resubmit the same script unchanged. A preview is a cheaper first check: avatar video previews generate the first-frame still stage so you can approve composition before spending a full render, though the docs say previews catch composition issues, not that they screen content. See also image moderation and content_policy_rejected.