Cybrlab: Preemptive AI shield against scams

[This article is part of a content series developed in collaboration with FEDIL, showcasing how artificial intelligence is contributing to the digital transformation of Luxembourg’s economy.]

Cybrlab.ai was founded in Luxembourg with the mission of "building a safer world by pushing the front lines of defence deeply". Although still a young company, its team brings together world-class engineers specialising in cybersecurity, resilient software systems, and artificial intelligence.

The scale of today’s cybersecurity challenges is staggering. In the past year alone, global losses from scams exceeded $1 trillion. This is a financial impact comparable to the combined damage of all climate-related disasters in 2024.

“You might wonder: for such a monumental problem, shouldn’t there already be an effective solution?” asks Garegin Margaryan, co-founder of CybrLab.ai. “The answer is no. If effective solutions existed, we would not be witnessing online scam activity growing seven times faster than the global adoption of the internet itself. The truth is that current defence approaches are fundamentally broken.”

The “Reaction Gap”

According to the CybrLab.ai team, the root cause is the deadly time lag between the onset of an attack and its detection. Modern scams are AI-enhanced, highly personalised, and executed at remarkable speed; 50% of attacks succeed within the first 24 hours. Yet conventional fraud detection systems analyse transactional data days later, typically trailing the actual theft by 5 to 10 days.

This problem is amplified by the fragmentation of digital platforms. A scam seldom occurs in isolation: it might begin with a WhatsApp message, escalate via a phone call, and conclude with a bank transfer. Because each platform sits in its own silo, even advanced defence systems fail to see the full picture.

With the rise of instant payments worldwide, recovering stolen funds has become nearly impossible. Meanwhile, EU and UK regulators are shifting liability by holding institutions accountable for scam-related losses. The need for an effective solution has therefore never been more urgent.

From 10 days to 5… minutes

CybrLab.ai is developing the ZeroScam ecosystem, which is far more than a typical fraud-detection app or Model-as-a-Service wrapper. It is a pre-emptive anti-scam environment designed to track the entire lifecycle of a scam.

It operates through four integrated pillars:

• The ZeroScam app – A free, anonymous mobile tool that allows individuals to verify suspicious messages. While protecting users in real time, it also crowdsources vital data for the detection engine.
• The ZeroScam engine – A sophisticated detection system comprising eight specialised modules.
• The ZeroScam API – A public interface for enterprises, giving financial institutions and telecoms actionable intelligence—such as flagged IBANs, crypto wallets, and malicious URLs—to block scams before funds leave a victim’s account.
• The Dynamic URL Scanner – A cornerstone of the ecosystem and a real-time threat-intelligence platform. Powered by a carefully engineered feature vector, in-house trained proprietary ML models, and world-class heuristic analysis, it is already proving to be one of the fastest and most accurate dynamic URL scanners on the global market.

When a user or an organisation uploads a screenshot or forwards a suspicious message through the mobile app or API, the system extracts the URL, analyses it, and delivers live threat intelligence within 30–40 seconds.

Immediate impact

This speed is transformative. During testing, already involving more than 10,000 real users across several countries, the system has successfully protected individuals against active attacks targeting major banks and government agencies.

By identifying fraudulent messages and flagging malicious URLs just minutes after a scam campaign begins, ZeroScam has prevented both financial loss for customers and reputational damage for the impersonated institutions.

CybrLab.ai’s advantage lies in its engineering pedigree. The team has previously built critical infrastructure and software from the ground up, successfully serving hundreds of millions of users.

The road ahead

“We are pushing the front lines of defence in depth by democratising cybersecurity,” says Garegin. “We combine the best open-source solutions with proprietary machine-learning models and algorithms to ensure that we’re not only reacting to attacks, but predicting them.”

CybrLab.ai plans to launch the production version of the ZeroScam ecosystem in early 2026. The company aims to protect up to 10 million people by the end of 2027 and is targeting more than 1,000 potential enterprise clients currently facing exposure to the global fraud epidemic.