African Startup Develops Machine That Detects Lies
In a corner of Africa’s rising tech landscape, an Ethiopian entrepreneur is attempting something extraordinary: to make truth measurable. Jibril Mohamed Ahmed, founder of the nonprofit startup Open Trust Intelligence, has developed what he calls the world’s first machine that detects lies — not through artificial intelligence, but through cryptography and deterministic algorithms. His innovation, which he names Trust Intelligence (TI), seeks to verify honesty in both digital and human interactions at a time when the world faces an overwhelming crisis of misinformation, deepfakes, and fraud.
Before creating TI, Jibril had already developed an AI image generator capable of producing highly realistic images. Reflecting on the potential impact, he realized, “It is only right, it is only fair if we give people the tools to detect what is real and what is not.” This insight led to the development of TruthLayer, which can detect AI-generated content from human-generated content with 98% accuracy.
Unlike conventional AI models that rely on massive datasets and probabilistic guesses, OpenTI works on the principle of verifiable computation. It measures consistency, coherence, and authenticity across text, audio, code, and video — without retaining or learning from personal data. As Jibril explains, “It’s designed to verify, not to learn. We don’t harvest, we don’t store, we don’t predict. We simply measure trust.”
Open Trust Intelligence powers two core systems. TruthLayer authenticates digital content, while Axumify analyzes voice, text, and video to assess potential deception and verify identities. Axumify can detect lies with 85–90% accuracy, compared to 54% for humans (American Psychological Association), and its verification system reduces fraud exposure in organizations by up to 70%. Together, these tools form what Jibril calls the “infrastructure of truth.”
The project was not born in a lab but over a simple coffee meeting. “I was having coffee with a close Danish friend who has been building Africa’s tech ecosystems for the past decade,” Jibril recalls, “and he suggested, ‘What if we had AI that could detect lies?’ Then he laughed and said, ‘That would be an ethical disaster — people might kill over it.’” Jibril smiled as he remembered. “He was right, but it made me think. The real disaster is that we already live in a world where no one can verify truth at all.” That conversation inspired Axumify, the second key component of OpenTI.
The urgency of his work is reflected in the numbers. According to a 2024 Regula report, 92% of global companies reported losses linked to deepfake-related fraud, with average damages of $450,000 per business. The UK’s Arup Group recently lost $25 million to a deepfake scam imitating a senior executive on video. In Africa, Smile ID reported a 200% increase in biometric fraud in 2024 alone, while MIT research shows false news spreads 70% faster than truth on social media. Jibril argues, “Trust has quietly become the rarest currency in the digital economy.”
Beyond technology, Jibril is pushing the idea of trust as an economic principle. In his academic paper, “The Economics of Trust Intelligence,” he presents a new theory that modern economics must include a fifth pillar — trust — alongside land, labor, capital, and entrepreneurship. “Without measurable trust,” he writes, “the foundation of every market, government, and institution eventually collapses.” Trust Intelligence, he argues, is not just a tool for verification, but an economic infrastructure for the 21st century.
OpenTI began as a public trial, attracting thousands of testers across industries and countries. “We gave it away to thousands for free — not for profit, but for proof,” Jibril said. “And the response was overwhelming.” Following feedback and refinement, access is now limited to institutional and premium users.
While OpenTI remains under evaluation, its implications are profound. In a world flooded with synthetic voices, fake faces, and algorithmic narratives, Jibril’s deterministic approach feels refreshingly human — a return to verifiable truth. “It’s not about outsmarting AI,” he says. “It’s about ensuring truth itself doesn’t go extinct.”