Working Group Charter: Standards for Islamic AI
Last updated 31 July 2026
Group name
Standards for Islamic AI
Mission
Answer one question with authority: for something to call itself an Islamic AI assistant, what criteria should it meet?
Deliverable
Version 0.1 of a published standard for Islamic AI assistants — versioned, versionable, and testable wherever possible.
The work starts with a definition workshop: is this a standard (measurable criteria) or a guidance document (what builders should consider)? The answer changes the deliverable, so it is settled first, not discovered late.
In scope
- The definition workshop, as the group’s first task.
- The criteria families: sourcing and citation requirements (does it cite real hadith? classical sources?); knowledge floors (benchmark thresholds); professed-values alignment; conduct under pressure (JaleesBench); disclosure requirements (madhab positioning, “not a mufti” boundaries, when to refer to a scholar); and data and privacy expectations.
- Conformance shape: possibly tiered (self-attested → benchmarked → reviewed) rather than binary.
- Drawing the measurable-criteria machinery from the Islamic Benchmarking group rather than rebuilding it.
Out of scope
- Anything beyond assistants, for v0.1. “Islamic AI applications” is a far bigger category — assistants-only is deliberate scope discipline, and the bar that keeps a standards group from becoming a debating society is a finishable deliverable.
- Building benchmarks. The Islamic Benchmarking group owns the measurement machinery; this group sets thresholds against it.
- Certifying products. v0.1 publishes the criteria; any certification or audit program is a later decision, made once the standard exists.
Duration, timeline, and milestones
Settle what it is, draft it, review it, publish it: the definition workshop first, then the criteria drafted family by family, then a review round with scholars and builders, then v0.1 ships.
T₀ = the chartering date (co-leads confirmed). Offsets firm up at the first milestone review.
| Milestone | What it concretely demonstrates | Review date |
|---|---|---|
| Definition settled | Workshop concluded: standard vs guidance decided, criteria families enumerated | T₀ + 6 weeks |
| Draft in review | Full v0.1 draft with testable criteria and thresholds, circulated to scholars and builders | T₀ + 14 weeks |
| v0.1 published | The standard live, versioned, with its revision process stated | T₀ + 20 weeks |
Resources needed
- Scholarly review of the criteria — several families (sourcing, disclosure, madhab positioning) are religious judgments.
- Benchmark thresholds and measurement machinery from the Islamic Benchmarking group.
- From IASER: a publishing home for the standard and editorial support. No compute to speak of.
Audience
Builders of Islamic AI assistants who want to make the claim credibly — with Ansari 4 as the natural reference implementation; and the users, communities, and institutions deciding which tools to trust, who read the standard as a checklist for what to ask of any assistant.
How we work
The standard lives in a new public repo under github.com/iaser-ai,
where versions and revisions are tracked in the open. Meeting cadence
set by the co-leads at chartering.