Table 2 Examples of what DIA adds to existing IAs

Scope

ISO sets organisation-wide AI governance/risk processes; FRIA triggers for high-risk and public uses; HUDERIA guidance for rights, democracy and rule-of-law risks

Applies to all stages of AI systems lifecycle irrespective of risk level especially when projects touch historically marginalised communities

Problem framing

Generic impact and risks; rights

Analysis of colonial legacies with explicit mapping of power imbalances and control

Epistemic justice

Not explicit

Requires pluriversal perspectives; knowledge sources and data labels

Consent/autonomy

FRIA/HUDERIA expect stakeholder engagement; consent often framed at data subject level

Adds collective/community consent

Data provenance & sovereignty (data and infrastructure)

ISO/FRIA check lawfulness, quality, bias; data sovereignty not required

Requires audits for coloniality of data (extractive practices, context loss), enforces data and infrastructure sovereignty (local storage, access terms) and culturally bounded data categories

Language and cultural rights

All have bias/fairness checks; no duty to support minority languages beyond performance metrics

Requires plans for linguistic inclusion, protection of cultural IP, and community control over cultural models

benefit-sharing

Not explicitly required

Requires equitable value-sharing (including local hiring/skills transfer) tied to design and deployment approvals

Procurement & local capacity

Supplier oversight and process controls (ISO 42001)

Encourages local suppliers or local partnerships; mandating training and institutional capacity-building as part of the innovation lifecycle