iGamingGenerator

Generator Prompt to Assemble Age Assurance Evidence for an eSafety Examination

This **Luna** generator prompt takes your production age assurance stack and Australia's **Online Safety Amendment (Strengthening Enforcement for the Social Media Minimum Age) Bill 2026**, then assembles the evidence pack an eSafety examiner now has the power to compel. Luna, the deepidv compliance overseer, builds a **method inventory** of every age assurance method in production and the buffer policies between them, compiles **measured performance** including accuracy and demographic-consistency figures on the actual user population, extracts **per-decision records** for a sampled week showing method, estimate, confidence, liveness verdict, fallback routing, and outcome, documents **circumvention handling** for borrowed-account, replayed-media, and false-age attempts, and issues a **gap statement** for any check whose records cannot be assembled automatically. Built for trust and safety leads whose enforcement regime just shifted from self-reporting to mandatory documentation disclosure, with penalties up to A$99 million.

How to use this prompt

1
Open Luna in the deepidv dashboard and paste the full prompt, or run it in Claude, ChatGPT, or Gemini if you are drafting the evidence-pack structure outside the platform.
2
Replace the INPUT section with your live age assurance methods, your measurement outputs, and a sampled week of decision records.
3
Run the prompt and read the gap statement first: it names any check whose records cannot be produced from system output on demand.
4
Route the method inventory and measured performance to compliance, and hand each gap to its owner with a remediation date.
5
Schedule the pack to refresh quarterly so it is always current rather than assembled under examination notice.

The prompt

Luna, assemble our age assurance evidence pack to eSafety examination standard, per the Online Safety Amendment (Strengthening Enforcement for the Social Media Minimum Age) Bill 2026.

Context: the amendment shifts enforcement from self-reporting to mandatory documentation disclosure, with penalties up to A$99 million. The examiner's question is not whether we have a program but whether we can produce its records.

Assemble, from system output only:
1. Method inventory: every age assurance method in production (facial age estimation, document verification, credential acceptance, inference signals), where each applies, and the buffer policies between them.
2. Measured performance: accuracy and demographic-consistency figures for the estimation layer on our actual user population, with the measurement methodology stated.
3. Decision records: for a sampled week, the per-check evidence trail: method, estimate and confidence, liveness verdict, fallback routing, and outcome, demonstrating reconstructability on demand.
4. Circumvention handling: detected borrowed-account, replayed-media, and false-age attempts, with the signal that caught each and the account action taken.
5. Gap statement: any check in production whose records could not be assembled automatically, flagged as remediation items with owners.

Deliver as a structured pack with an executive summary, and schedule a quarterly refresh so the pack is always current rather than assembled under notice.

Test it in Claude or another LLM

This prompt is built for the Luna agent inside deepidv, where Luna reads a platform's live age-verdict records and measurement outputs. You can dry-run the pack structure in any general LLM first with synthetic data before pointing it at real records.

1
Paste the full prompt into Claude, ChatGPT, or Gemini, but replace the opening 'Luna,' with a role instruction such as 'Act as an age assurance evidence assembler preparing for an Australian eSafety examination.' Keep the OUTPUT sections exactly as written.
2
Under the INPUT section, paste the synthetic sample block below so the model has methods, measurement figures, and sample decision records.
3
Add one framing line: 'This is synthetic test data. Where a record cannot be assembled from the input, flag it in the gap statement instead of inventing it.'
4
Check the output shape: a method inventory, measured performance, per-decision records, circumvention handling, and a gap statement, with an executive summary. If any figure is invented, tighten the framing line and re-run.
5
Once the output shape is right, run it live in the deepidv dashboard where Luna assembles the pack from your real system output.

Synthetic sample data to paste alongside the prompt

Fake test data, safe to share with any LLM. Swap in your own once the output looks right.

AGE METHODS (synthetic, fake): facial age estimation over passive liveness, document verification fallback, mDL/ZKP acceptance, inference signals; buffer set to estimated 21+ for an 18+ gate
MEASURED PERFORMANCE (fake): MAE 2.3 years overall; per-demographic bands provided; methodology note attached
DECISION RECORDS (fake): one sampled week, ~10k checks with method, estimate, confidence, liveness verdict, fallback, outcome
CIRCUMVENTION LOG (fake): borrowed-account and replayed-media attempts with catching signal and account action
OPEN ITEM (fake): one legacy web flow whose per-decision records are not automatically retained

Pairs with on deepidv

Sources & further reading

FAQ

What does an eSafety examination now require?

Australia's amendment shifted enforcement from platform self-reporting to mandatory documentation disclosure, with penalties up to A$99 million. An examiner can compel the records behind a program: which age methods run, their measured accuracy on the real population, the decision trail for contested accounts, and how circumvention is handled.

What does this prompt assemble?

A structured evidence pack: a method inventory with buffer policies, measured accuracy and demographic-consistency figures, per-decision records for a sampled week, circumvention handling with the signal that caught each attempt, and a gap statement for any check that cannot be reconstructed from system output.

Can I run this outside the deepidv dashboard?

Yes. The structure works in Claude, ChatGPT, or Gemini as an evidence-pack framework. Assembling the pack from your real age-verdict records, measured accuracy, and circumvention logs only runs when it executes inside the deepidv dashboard through Luna.

Run it with live verification data

These prompts work in any LLM. Inside the deepidv dashboard, Luna, Arbiter, and Arc run them against your real sessions, screening lists, and audit trails.

Book a Demo