AI in election processes: The dangers of uncertified and proprietary solutions, by Chris Uwaje and Yakub Aliyu

MEMORANDUM TO INEC NIGERIA:

Attention: Chairman, Independent National Electoral Commission (INEC), National Information Technology Development Agency (NITDA), Office of the National Security Adviser (ONSA), African Union – Political Affairs, Peace And Security.

SUBJECT: “Urgent Concerns on: Deployment of Uncertified and Proprietary AI-Playbooks for National Election Verification Without Certified National AI Solutions Providers.

Mandate: Deliver Open, Sovereign, and Certified Electoral Systems.”

FOCUS: Critical AI Side-Effects in National  Electoral Processes and National Digital Sovereignty.

EXECUTIVE SUMMARY.

OUR CONCERN:

AI-AI EVERYWHERE AND NOTHING TO VERIFY? WHAT WAY FORWARD FOR NIGERIA NATIONAL GENERAL ELECTION 2027?

And the Oldman warner by asking: “Did our forefathers fight colonialism to be ruled by a Dashboard whose password we do not know, not verifiable nor, controllable?”

Great question to a conscious nation!

Meanwhile, without any doubt of ambiguity, tested AI Solutions requires an end-to-end Internet Protocol version six  (IPv6) infrastructure to scale. Meanwhile our critical digital infrastructure are currently predominantly driven by Internet Protocol version four IPv4/NAT. This is not AI configured nor operationally friendly.

Perhaps, we are about to witness a silent digital coup orchestrated by the insertion and invasion of proprietary AI-black-box as the anticipated new digital Commander of a playbook to deliver a delusional election collation, voters accreditation, and result verification pipelines – imported by vendors who are not accedited nor certified by any Nigerian Software Practitioners and regulatory body? More worrisome is the fact that it is not feasible to effectively deliver the mission within a 3 mount tineline.

The preceeding position and related

questions become imperative to interrogate the recent pronouncement by INEC with regards to the delivery of AI-centric solution for the upcoming 2027 National General Election.

This Memo sets out to interrogate the national security threat and the anticipated disastrous effects of deployment of AI as an umpire in a National Election – particularly in a digitally underdeveloped atmosphere?

Above all, it calls on INEC for clarifications on the late-hour declaration and directives on the holistic deployment of AI to govern the scheduled 2027 electoral process without Certified, tested, Open-Auditable, and Sovereign AI Solutions. Is the messaging confirming that the nation has outsourced her electoral process deviod of secured and mandatory digital infrastructure governaned by Internet Protocol version six (IPv6)?

An election verified by uncredited and uncertified, foreign-Agents owned proprietary AI is not verifiable but can only represent a covert Agentic outsourced mission!

1.0 THE PROBLEM: THE CURRENT AI INTERVENTION PLAYBOOK

Available records indicate that African Election Management Bodies (EMBs) are currently buying AI as a showcase of “feature” signposts with the following mindset agenda:

– Voter register de-duplication of available software.

– BVAS facial recognition module

– IReV image upload and readability AI

– Result collation anomaly detection

– Social media disinformation monitoring.

And whereas, those AI modules are:

1.1. Uncertified: Not tested or licensed by NITDA, NDPC, or Standards Organisation of Nigeria (SON) as meeting the Nigerian AI Standard. Indeed, No one knows their false positive rate of images on dark-skinned faces, or their error rate on handwritten Arabic or Hausa and other local language numerals.

1.2. Proprietary: The intent of AI vendor is to sell a product based on the slogan: “Trust our AI.” But that is not science. Rather, at best the claim can only be classified as ‘a priesthood rendition’. Therefore, concerns on Closed Source code becomes magnified and a deeper digital sovereignty concern.

For example. indications suggest the possibilities of INEC’s access to and ability to the backend view of how a result was flagged as “fake”.

1.3. Playbook without Accountability:

But indeed, accountability demands authentication, but AI electoral process playbook has no signature! In this regards, can AI be called and admitted as an authentic witness in the Court of law?

The AI Machine tells us: If AI flags result, quarantine PU. But who wrote that rule? Who owns liability if AI wrongly cancels 500,000 votes?

2.0 IMPLICATIONS OF PLUGGING INTO THIS AI-DEEP-HOLE:

2.1 Sovereignty Implosion:

If the AI that verifies elections is proprietary and foreign-hosted, the owner of that AI can, with a single update, change the outcome of anomaly detection without entering Nigeria. Does that mean that indeed, we may have moved from ballot stuffing to sophisticated algorithm stuffing? Especially when the machine cannot be identified?

2.2 Legal Nullity:

Without being a Lawyer, every literate layman knows that; Under Section 148 of Electoral Act 2022, any party can demand forensic inspection of election materials. But the irony is that you cannot forensically inspect a proprietary black box that says “Trade Secret.” This may therefore mean that through the inclusion of AI advisory,

we have clouded the constitutional crisis into a domain of digital complexity! Defence Lawyers will enter a shouting match of “SHOW ME THE KAW”! This  is because, any election decided by such AI is liable to be nullified by the Tribunal if challenged.

2.3 Tribalization of Error: Another critical factor is the Tribe identification and classification error.

This increases the complexities of election conflicts, since uncertified AI trained on Caucasian faces fails 3 times more on African women and youth. If such AI is used for BVAS accreditation, it will disproportionately reject women in purdah, farmers with worn fingerprints, as well as young first-time voters. Classically, that is not a bug. But outright voter suppression by digital code.

2.4 Vendor Lock-in and Extortion

Once INEC builds its entire verification history on a proprietary AI database format, that vendor can increase license fees by 1000% before next election. That means that: ‘You cannot migrate. Your national election history is held hostage’. This has happened in 3 African countries since 2023.

2.5 The Deep Flaw;- Kill Switch Risk:

Proprietary AI requires Internet call-back to foreign license server. In a contested election, that server can be “down for maintenance.” That means that the verification dies. It is emphatic to note that a certified, on-premise, open system does not have a kill switch.

3.0 THE CALL MANDATE: BAN UNCERTIFIED PROPRIETARY AI IN NATIONAL  ELECTION 2007.

The Election AI Sovereignty Doctrine for adoption by AU and INEC:

DOCTRINE 1: CERTIFICATION IS NON-NEGOTIABLE: No AI module shall be used in any stage of election (registration, accreditation, voting, collation, verification, litigation) unless:

– Certified by NITDA + SON under the Nigeria AI Assurance Framework

– Tested on Nigerian dataset (100,000+ handwritten EC8A forms, 500,000+ Nigerian faces across 6 geo-zones, 4 Nigeruan languages)

– To Published Accuracy, Bias, and Test-Failure Report (ABF Report) 90 days before election

DOCTRINE 2: NO PROPRIETARY BLACK BOX IN VERIFICATION:

All AI used for verification must be:

– Open-Source Auditable: Code deposited with Federal High Court and Nigerian Bar Association.

– On-Premise & IPv6-Native: No foreign cloud call-back. Must Run on Nigerian Servers, and on Nigerian IP addresses.

– Human-Overridable: Any AI flag must be reviewed by 2 humans – One each from INEC tech and party agent. With access view on a split screen showing: why it flagged. No “89% fake” — must show pixel proof.

DOCTRINE 3: CERTIFIED AI SOLUTIONS PROVIDERS POOL:

Create a pre-qualified pool of Certified Nigerian AI Solutions Providers (CN-AISP). Criteria:

– 51% Nigerian-owned, 30% youth and women in technical roles

– Office and servers reside in Nigeria

– Willing to train 1000 NYSC corps members per election cycle in AI forensics.

– Code escrow agreement with the state.

Assurances that INEC shall ONLY procure election AI from the above pool.

DOCTRINE 4: THE BALANCE VERIFICATION PLAYBOOK:

Replace vendor playbook with Concerned Citizens (CC) Playbook:

– Flag, Don’t Cancel: AI flags anomaly Endure that Human Team goes to Pooling Unit to conduct physical verification before final decision.

– Annomaly Dual-AI Consensus: No single AI decides. One open-source Nigerian AI and one open-source AU AI must both agree there is an anomaly before resolution.

4.0 ACT NOW: IMMEDIATE ACTIONS:

4.1.  INEC to issue moratorium on renewal of any proprietary AI election license until the completion of officual Audit.

4.2.  NITDA to publish within 30 days the Certification Checklist for Election AI.

4.3.  AU to adopt this memo as Continental Guideline for 2027-2028 election cycle.

5.0 CONCLUSION

To all Great-Minds out there, history will not remember who coded BVAS. It will remember who owned the code that verified the will of 230 million people.

If we allow proprietary, uncertified AI Solutions to be the judge of our elections, we have replaced human jury with AI jury in the court of the people of Nigeria  – the most important court for the determination of freedom of choice and trust of our vote.

Final word: Therefore, the 2027 electoral processes demand that the illusion on AI electoral process calls for: “Nothing on our electoral votes, without IPv6 and our human codes.”

From Concerned Citizens of Africa AI-IPv6-Cybersecurity Task Force. AI. Promoters: Chris Uwaje, and Prof.  Yakub Aliyu (USA).

Released On October 7, 2026.