An AI Design Workshop Opens Today: A Governed Cultural-Content Workflow Linking Canva, AIGC and Podcasts for Taiwan’s 55 Indigenous Townships

The Indigenous Peoples Cultural Development Center of the Council of Indigenous Peoples, Taiwan begins its “Visual Beauty, Cultural Appeal: Cultural Marketing and Content Creation Workshop” series today, August 24. The official program links three capabilities in sequence: the first session introduces Canva visual design and explicitly includes AI design; September sessions move into social-media operation and back-end analytics; and an October session focuses on podcast storytelling and script development. See Indigenous Peoples Cultural Development Center | Visual Beauty, Cultural Appeal.

For township offices, community organizations, cultural institutions, youth teams and local brands, the useful lesson is not simply that more AI tools are available. The more durable opportunity is to make every design, social post and podcast episode return to the same governed resource layer. A poster, an Indigenous-language recording, a photograph related to ceremony or a podcast episode can then retain provenance, permissions, cultural context, version history, publication limits, AIGC permissions and a practical withdrawal route instead of becoming a scattered file after an event ends.

Layer 1: Build a “cultural resource passport” before scaling generation

Start by assigning every resource a stable ID. A minimum record can include `resource_id`, `people_or_community`, `content_type`, `original_provider`, `editor`, `cultural_reviewer`, `created_at`, `version`, `source`, `access_level`, `permitted_uses`, `aigc_derivative_use`, `model_training_permission`, `last_reviewed_at`, and `withdrawal_contact`. Images can add consent and location restrictions; audio can add speaker, language or variety, editing permission and transcription permission. Ceremony, family memory, traditional territory, gathering practices and other sensitive knowledge can trigger additional community-governance fields.

The point is not to create paperwork. It is to reduce how much later systems have to guess. An AI visual-storytelling tool can directly filter for “public, reviewed, social-card reuse permitted.” A podcast team can verify whether a historical recording may be excerpted before editing. Incomplete records can stay in a normal pending review state rather than allowing AIGC to infer identity, community, place or cultural meaning.

UNESCO’s discussion of ethical generative-AI use of Indigenous data notes both the potential of AI for recording, transmission and revitalization and the importance of Indigenous data sovereignty, cultural context and protection against misuse. A resource passport turns those governance concerns into fields a local system can actually enforce.

Layer 2: Keep AI in the drafting role and cultural authority with people and communities

Canva and other generative tools can rapidly create layouts, text drafts, visual concepts and short-video scripts. Formal publication should nevertheless pass through identifiable human review. A practical workflow is:

`original resource → AI-assisted draft → language/factual review → cultural-permission review when needed → publishable version`.

Routine public information can be reviewed by the responsible editor or administrator. Material involving Indigenous languages, ceremony, traditional knowledge, family narratives or sacred content can trigger review by an appropriate knowledge holder, language educator or community process.

UNESCO’s 2026 AI and culture dialogue emphasized that generative AI is reshaping cultural production and distribution and that validation, human intention, originality and accountability remain central. For Indigenous cultural content, that distinction is highly practical: AI can accelerate routine production, while people and communities retain the authority to decide what is culturally appropriate and what may be shared.

Layer 3: Let AI visual stories, social posts and podcasts share one master record

The workshop series starts with visual design and later connects social media and podcasts. That sequence can become a one master record, many outputs workflow. A single event record can support a website story, Instagram or Facebook card, short-video subtitles, podcast show notes and chatbot answers. Each derivative retains its source resource ID and version.

If a name is corrected, an image is replaced, a participant withdraws consent or a cultural-use condition changes, the master record can identify every outlet that must be updated. This is useful for an AI visual journal and a 55-township public-information index as well: AI may assist with cropping, summaries, captions and layouts, but every operation first reads the resource’s permitted use.

A resource marked `aigc_derivative_use=false` should not be sent into a generative workflow. A record marked community-only should not be transferred to a public model. Encoding those rules in the data layer is more reliable than repeatedly asking a model in a prompt not to disclose or transform restricted content.

Layer 4: RAG should be permission-first, retrieval-second and generation-last

If cultural resources are later connected to a chatbot or a Two-Eyed Seeing knowledge lab, RAG should not place every record into one vector store and try to hide sensitive material afterward. A more robust sequence checks user role, purpose and resource permission first; retrieves only allowed records; and only then generates an answer. Responses can display provenance, version and last-review date.

Withdrawal must propagate beyond the visible web page. It should invalidate vector chunks, caches and pre-generated summaries as well. The Global Indigenous Data Alliance’s CARE Principles—Collective Benefit, Authority to Control, Responsibility and Ethics—offer a useful governance test for this architecture. Local Contexts’ Traditional Knowledge Labels can also inform metadata for sacred or ceremonial restrictions, seasonal conditions, gender-specific protocols, non-commercial use and other community-defined rules.

Layer 5: Traditional cultural expressions should be creatable and protectable

Successful cultural communication increases the likelihood that material will be seen, copied, remixed and reused through generative tools. WIPO treats songs, dances, designs, ceremonies, stories, crafts and other intergenerational expressions as part of the field of Traditional Cultural Expressions. See WIPO | Traditional Cultural Expressions.

For a local cultural-content workflow, one useful default is that publicly viewable does not mean available for AI training. A community or creator can separately decide whether a resource may be viewed publicly, quoted for education, shared socially, used commercially, transformed by AIGC or used in model training. Those are different permissions.

This approach supports cultural circulation rather than blocking it. Creators and public-service teams gain a clearer map of which resources can be reused confidently and which require renewed consultation.

A 90-day MVP: organize 30–50 low-risk resources before building a large platform

A township office or cultural team does not need a large platform to start. During the first 30 days, select 30–50 high-use resources with clear provenance and relatively low cultural risk, then establish their passports and permissions. During days 31–60, build a simple search interface, AI visual-storytelling prototype or small RAG test that exposes only reviewed and explicitly permitted records. During days 61–90, invite administrators, youth, language educators, cultural workers and knowledge holders to test misclassification, provenance loss, cultural labeling, withdrawal and synchronization across channels.

Useful measures include the share of resources with complete provenance, the percentage with explicit permissions, human-review completion, time from correction to synchronization, successful withdrawal from search indexes, ability to locate the current version quickly, and whether AI answers visibly cite their sources. Social reach still matters, but it does not need to be the only measure of public value.

From “learning AI” to retaining local capability

Today’s Canva and AI design session is a practical entry point. The later social-data and podcast sessions connect visual communication, audience understanding and sound storytelling. If Taiwan’s 55 Indigenous townships add a governance layer, one-off outputs can become durable local capacity: every image, recording and post can retain where it came from, who has authority over it, what uses are permitted, when it must be reviewed and how it can be withdrawn.

That foundation can later support public-notice indexing, a chatbot MVP, a Two-Eyed Seeing knowledge lab, AI visual storytelling and RAG without rebuilding cultural rules for every tool. Technology does not need to replace cultural practitioners. Its stronger role is to reduce repetitive organization while keeping narrative authority, permissions and cultural judgment with communities.

Sources

• Indigenous Peoples Cultural Development Center | Visual Beauty, Cultural Appeal
• UNESCO | Leveraging UNESCO Normative Instruments for an Ethical Generative AI Use of Indigenous Data
• UNESCO | AI and culture
• WIPO | Traditional Cultural Expressions
• Global Indigenous Data Alliance | CARE Principles
• Local Contexts | Traditional Knowledge Labels