Account policies

Each account carries a set of policies that tell the pipeline how to behave for that account — how to source images and video, when to require captions, whether to label AI content, how to engage with audiences, and how to run promotion. You edit them from the account page; each policy opens as its own form. The account’s CustomerAdministrator, an assigned AccountOperator and platform admins can edit them; a CustomerViewer sees them read-only.

🎬 Set an account policy

Prefer to read? Open the step-by-step transcript

1. Open the account, then pick a policy card (Accessibility, Image Acquisition, …).
2. Adjust the settings — most policies have a Policy active switch plus their own fields.
3. Save. The change applies to the account’s next productions.

The policies at a glance

Every account shows all ten policy cards, in alphabetical order. The last two only matter for accounts that use the matching theme.

Policy	Governs	Key controls
Accessibility	WCAG conformance of deliverables	Standard (WCAG 2.1 AA/AAA, 508, EN 301 549), require captions / audio-description / transcript, audio-description timing, human review of audio descriptions
Image acquisition	How images are sourced	Strategy (Stock / AI / Operator / Hybrid), stock library priority, safety + alignment thresholds, copyright-risk auto-reject
Video acquisition	How footage is sourced	Strategy (Stock / AI text-to-video / image-to-video / Broadcast feed / Hybrid), library priority, avatar animation
Content risk	Brand-safety guardrails	Flag public figures / logos / mature content, allow identifiable real people (adds a consent gate)
Content disclosure	AI-authorship labelling	Require disclosure, text, position (intro / outro / overlay), format (text card / spoken / both), what it applies to
Financial services	FCA compliance for regulated content	FCA audience classification, mandatory pre-publish review, financial-content guardrail
Promotion	Loop 5 paid + organic promotion	Enable paid campaigns, budget caps, promotion window, monetisation metric, platform scope
Engagement	Loop 3 audience interaction	Which interaction types to answer, tone, auto-post vs review, daily limits, AI-reply disclosure
Talks Policy (Talks accounts)	The look + rules of Talks debates	Speaker balance, composition style (split-screen / alternating / reaction), consistency + coherence scores, caption format
Historical Animation (Time-Travel accounts)	Animating archival imagery	Image source (AI / Wikimedia), colourisation, motion effect, animation focus

What each policy does

Accessibility

Sets the accessibility standard to target and whether deliverables must include captions, audio descriptions and transcripts — each toggle activates the matching pipeline step. The standard you choose is recorded on the production’s accessibility gate result, so an audit can see what each production was held to.

Audio-description timing sets how long each description may be: Standard keeps it to a sentence or two that fits the natural pauses in the narration; Extended describes the scene fully, for playback that pauses the video. You can also require human review of audio descriptions before the production clears its gate. See the platform Accessibility statement.

Note

Captions are written from the production’s own script, so there is no caption accuracy to set — the form has no accuracy threshold.

Image acquisition & Video acquisition

Choose the sourcing strategy — stock libraries, AI generation, operator-supplied, or a hybrid — and the priority order of stock libraries. Acquisition enforces a minimum content-safety and prompt-alignment score, can auto-reject high copyright-risk images, and falls back to AI generation when allowed. Video acquisition adds clip-length and cost caps and an avatar animation option.

Content risk

Brand-safety guardrails: flag identifiable public figures, brand logos and mature content. Allowing identifiable real people is off by default and, when enabled, adds a consent/approval gate before publish.

Content disclosure

Controls whether productions carry an AI-authorship disclosure, and if so the text, position (intro, outro, or persistent overlay) and format (on-screen card, spoken, or both) — relevant to FTC guidance and EU AI Act Article 50.

Financial services

For regulated finance content, set the FCA audience classification. When it’s anything other than Not applicable, you can require a mandatory compliance review before distribution and keep the financial-content guardrail on (strongly recommended).

Promotion

Governs Loop 5. With paid promotion off, JARAI promotes organically only. Turn it on to set a budget cap per production, a promotion window, the monetisation metric (views, watch-time, streams, revenue), and an optional platform scope and per-platform budget overrides.

Engagement

Governs Loop 3. Pick which interaction types to respond to, the tone, and whether replies auto-post or wait for operator review. Daily caps on responses and sentiment analyses keep spend bounded, and you can append an AI-reply disclosure.

Talks Policy & Historical Animation

These cards show on every account but only take effect when the account uses the matching theme — a Talks account needs an active Talks Policy before its Talks video step can run. Talks Policy sets speaker balance, the composition style (split-screen, alternating, reaction shots), quality thresholds and caption format — see Talks queue. Historical Animation sets the archival-image source (AI or Wikimedia, which requires attribution), colourisation, and the motion style.

Note

Most policies have a Policy active switch (Engagement’s is Engagement policy active, which suspends Loop 3 for the account when off; Promotion has none — its Enable paid promotion campaigns switch decides between paid and organic-only). Turn a policy off and JARAI falls back to safe platform defaults (e.g. all content-risk flags on, real-person content disallowed). Policy changes apply to the account’s next productions, not ones already running.

Tip

A self-service account’s own CustomerAdministrator can change its policies — they don’t need an operator to do it. A CustomerViewer can see them but not change them.