In Nigerian general elections, the most critical manipulation frequently occurs away from the capital city in rural polling stations and ward collation centres, often protected by political thugs. When aggrieved political candidates attempt to present mobile phone footage at election petition tribunals, they encounter significant legal challenges under Section 84 of the Evidence Act and the growing prevalence of generative artificial intelligence, which defense lawyers use to dismiss genuine recordings as digital fabrications.
To address this evidentiary vulnerability ahead of the 2027 polls, Jason Omu, a finance professional turned infrastructure developer and co-founder of the civic-tech platform Pawavox, has developed a cryptographic verification protocol. Incubated at Fast Forward Venture Studio with support from Prosperity Tech, the system evolved from a basic feedback tool into a specialized security architecture designed for high-risk environments.
“Now, Pawavox stands as a way to give data a stronghold in court,” Omu explained. “Whenever someone information is presented, it is no longer a question of whether it is real, AI-generated, or manipulated. Once recorded on Pawavox, traditional fact-checking is bypassed because altering even a single second of video or adjusting the brightness will break the cryptographic hash.”
What is Section 84 and why does it affect digital evidence?
Under Section 84 of the Evidence Act, electronic recordings face stringent scrutiny regarding device integrity and chain of custody, frequently undermining viral polling unit videos alongside official Independent National Electoral Commission ([INEC](https://technext24.com/news/inec-probes- suspicious-voter-register-ai-photo/)) Bimodal Voter Accreditation System ([BVAS](https://technext24.com/news/inec-strengthen-bvas-software-introduce- ai/)) records. Pawavox addresses this by embedding four distinct cryptographic proofs into every captured file: a mathematical hash, the unique hardware identifier of the mobile device, a precise timestamp, and exact geographical coordinates designed to withstand tribunal cross-examination.
To prevent malicious actors from recording deepfakes off high-resolution screens or spoofing their positions via Virtual Private Networks, the application incorporates a rigorous analysis layer. While standard VPNs often spoof locations within a broad decimal range, Pawavox records coordinates up to 25 decimal places, immediately flagging discrepancies and running secondary checks to detect screen-played footage.
How does the app protect field observers in hostile zones?
Beyond data integrity, physical security remains a major concern for observers operating in volatile areas where ballot boxes are frequently targeted. Pawavox features an encrypted local vault that isolates captured media from the smartphone's standard gallery, ensuring that casual inspections by hostile actors reveal an empty camera roll. In extreme situations, observers are advised to rely on encrypted voice notes rather than video to safeguard their wellbeing.
If a device is permanently seized and exploited to upload fraudulent reports, administrative backend tools can identify the outlier hardware ID and quarantine its submissions from the broader dataset. Looking toward the 2031 electoral cycle, the startup aims to incorporate discreet wearable hardware, such as smart glasses and body cameras, to remove smartphones from hostile zones entirely.
How does Pawavox handle low-connectivity regions?
Operating across Nigeria's fragile telecommunications infrastructure presents notable connectivity hurdles, particularly in rural communities relying on congested 2G and 3G networks. Pawavox addresses this by splitting its data pipeline: lightweight signed JSON metadata is transmitted via the Nostr protocol, while heavy video payloads are routed separately to Google Cloud. The cryptographic signing occurs entirely offline, allowing data to sync in stages once a minimal network signal becomes available.
How does ElectionGPT process local languages without hallucinations?
To help situation rooms analyze thousands of concurrent submissions without manual overload, the team built a specialized chat interface known as ElectionGPT. Unlike open-web large language models prone to generating false information, ElectionGPT is completely offline and strictly restricted to verified field databases.
To manage the linguistic diversity of Nigerian polling units, where citizens frequently speak Yoruba, Igbo, Hausa, and Nigerian Pidgin, the platform integrates local language datasets supported by YarnGPT. When background noise or regional dialects reduce transcription certainty, the system assigns a confidence score rather than guessing, prompting human analysts to review the attached audio file directly.
What are the scaling plans beyond 2027?
During its initial test run at the Osun State election, the startup successfully onboarded 300 users and captured 142 verified reports, highlighting the challenge of scaling across Nigeria's 176,974 polling units. For the upcoming general elections, Pawavox is deploying a ₦13 million grant from the AI in Democracy Fund (AIDF) Sandbox to train an independent network of 5,000 observers while simultaneously pitching established civil society organizations to integrate its verification layer.
Looking past the electoral cycle, Omu envisions Pawavox operating as a hybrid social enterprise that combines public civic infrastructure with a sustainable B2B Software-as-a-Service model, offering verified field data solutions to corporate monitoring and evaluation teams across Africa and beyond.