RESPONSIBLE AI SCORECARD — REVISION 1.1
AGENTPR™ Responsible AI Scorecard
A public record of AGENTPR™'s current responsible AI posture, assessed against Microsoft RAI Standard v2, NIST AI RMF 1.0, EU AI Act, and NDPA Nigeria 2023. Updated as evidence accumulates.
THE ORIGINATING FRAMEWORK
Every commitment on this scorecard originates from the Engineering of Trust™ framework — developed by Dr. Celestine N. Achi as part of The Achi Intellectual Architecture™. The Engineering of Trust™ framework has been cross-walked against relevant international responsible-AI and data-governance standards to assess alignment, identify gaps and guide continuous improvement.
A standards cross-walk is an alignment exercise. It does not constitute certification, regulatory approval or endorsement by Microsoft, NIST, the European Union or the Nigeria Data Protection Commission.
Version 1.1 clarifies the distinction between evidence provenance, claim confidence, task-level model confidence and human-review status; corrects standards-cross-walk terminology; and updates the Meaning Intelligence research status.
How AGENTPR™ communicates analytical certainty
Evidence provenance, claim confidence, model confidence and human-review status answer different questions.
| Assurance layer | Question answered | AGENTPR™ expression |
|---|---|---|
| Evidence provenance | Where did the evidence come from? | Monitored, Web-Verified, Print-Verified, AI-Sourced, AI-Inferred |
| Claim confidence | How strongly is the claim supported? | High, Medium, Low |
| Model confidence | How certain was the model about a defined machine task? | Percentage or task score, where applicable |
| Human review | Must professional judgement intervene? | Standard Review, Human Review Recommended, Human Review Required |
These layers operate together but are not interchangeable. A high model-confidence score does not convert an AI-sourced item into verified evidence and does not override a human-review requirement.
Model confidence is a machine-generated task score. It is not evidence verification, claim confidence or permission to publish. A high model-confidence score must never bypass required human review.
Implementation example
GASKI election intelligence
In the GASKI election-intelligence implementation, AI-assisted polling-unit sheet transcription is separated from publication authority. Machine-read records carry task-level confidence scores and remain pending until an authorised human reviewer verifies or rejects them against the original INEC sheet. Discrepancy flags indicate records requiring review; they are not presented as proof of wrongdoing.
GASKI reports figures attributed to INEC and does not declare election results.
RAI Dimension Assessment
Engineering of Trust™
Originating Framework — The Achi Intellectual Architecture™
The proprietary framework that preceded and informed this cross-walk. Methodology, Transparency, and Accountability at every stage of the GASD™ pipeline.
Transparency
T1 — Microsoft RAI Standard v2
Three claim-confidence levels are supported by five evidence-provenance labels, task-specific model-confidence scores where applicable, and a separate human-review status. Source attribution throughout. Glassbox AI Policy published. Live calibration data collected.
Inclusiveness
I1 — Microsoft RAI Standard v2
Meaning Intelligence™ is calibrated for Nigerian sarcasm, Pidgin English, and African institutional context. Initial preprint published; expanded benchmark validation and continuous calibration ongoing.
Individual Disclosure
T2-T3 — EU AI Act Art. 50 / NDPA 2023
Transparency Register published. Disclosure modal gates every brief generation. Audit trail in Supabase.
Contestability
F2 — NIST AI RMF 1.0 / EU AI Act
Self-service contestation portal live. Reference numbers issued automatically. Human review within 14 business days.
Human Oversight
A5 — Microsoft RAI Standard v2
Human Review Recommended is a workflow status for uncertain or high-risk outputs. Human Review Required blocks publication where authorised verification is mandatory. Feedback loops collect calibration data.
Measurement
Measurement — NIST AI RMF Measure 2.5
Feedback loop deployed June 2026. Calibration report live and auto-updating. Minimum threshold: 50 submissions. Full report publishes on threshold crossing.
Security
PS2 — Microsoft RAI Standard v2
Role-gated access, logged internally. Prompt-injection hardening and OAuth scope audit in active roadmap.
Active Commitments
What we are building and by when. This roadmap is the documented path to the next full RAI assessment.
PRIORITY 1
0–90 Days
- Publish DPIA summary
- Publish sarcasm evaluation methodology
- ZALI system card
- Sensitive use review for government tenants
- Calibration report — 50 feedback threshold ✓ Live
PRIORITY 2
90–180 Days
- Quarterly error-rate parity report
- Contestability workflow formal build ✓ Live
- Prompt-injection hardening
- OAuth scope audit for MCP
- GDPR addendum for EU-resident subjects
- C2PA content credentials on Narrator
PRIORITY 3
180–365 Days
- Independent third-party RAI audit
- ISO/IEC 42001 readiness assessment
- Red-team programme for Narrator
- Public incident log
- Expanded Meaning Intelligence™ benchmark validation
- RAI review board — quarterly cadence
Standards cross-walk
The Engineering of Trust™ framework has been cross-walked against relevant international responsible-AI and data-governance standards to assess alignment, identify gaps and guide continuous improvement.
A standards cross-walk is an alignment exercise. It does not constitute certification, regulatory approval or endorsement by Microsoft, NIST, the European Union or the Nigeria Data Protection Commission.
An internal standards cross-walk was completed in June 2026. Independent assurance remains on the published Responsible AI roadmap.
Governance
| Accountable Executive | Dr. Celestine N. Achi — Founder & CEO, Cihan Media Communications |
|---|---|
| Technical Co-owner | Orimolade Oluwamuyemi, FIIM — Strategic Technology Adviser |
| Relevant Nigerian Data Protection Regulator | Nigeria Data Protection Commission (NDPC) |
| Scorecard Revision | v1.1 — September 2026 |
| Assessment Baseline | v1.0 — June 2026 |
| Next Full Assessment | December 2026 |
| Scorecard Published | useagentpr.com/responsible-ai/scorecard |
This identifies the relevant national regulatory authority. It does not imply that the NDPC has approved, certified or endorsed AGENTPR™.
Initial preprint published; expanded benchmark validation and continuous calibration ongoing. Read the June 2026 preprint.
Responsible AI is not a destination. It is a cadence.
AGENTPR™ publishes this scorecard because the Engineering of Trust™ framework requires accountability to be visible, not assumed. The gaps are documented. The roadmap is live. The next assessment is scheduled.