Amid growing concerns over AI-generated misinformation and hallucinated outputs, Matthijs de Vries, founder of data infrastructure firm Nuklai, argues that better model architecture alone is insufficient.

Speaking with Benzinga, de Vries said the deeper issue lies in fragmented and unreliable access to real-time, verifiable data.

"I don't believe hallucinations are caused solely by poor data infrastructure, that would be an oversimplification," de Vries said. "That said, insufficient or fragmented access to real-time, verifiable data makes the problem significantly worse."

Nexus, which launched this week with backing from Filecoin (CRYPTO: FIL), io.net (CRYPTO: IO), and Fetch.ai (CRYPTO: FET), introduces a structured approach to data retrieval by using a virtual SQL interface to unify interactions across databases, APIs, files, and even non-SQL data sources.

It is designed to give AI systems a consistent way to access live, traceable information, reducing the likelihood of fabricated responses.

Unlike other protocols that rely heavily on embedding large datasets or using isolated tool chains, Nexus uses metadata-driven orchestration and SQL-like logic to fetch context-specific data only when needed.

This reduces latency, limits confusion between tools, and allows AI systems to present transparent answers linked to their original sources.

"Too many tools just sound intelligent on the surface but fall apart when it counts," de Vries said. "With Nexus, we're changing that. It connects straight to the data, shows you where answers come from, and finally fixes the hallucination issue everyone's been dealing with."

The product launch also highlights the broader architecture of Nuklai's ecosystem, which includes Helix, its blockchain-based layer for value coordination and data governance.

De Vries emphasized that all significant interactions, including collaborations between partners or developers, will be traceable and governed on-chain.

He also addressed trade-offs in integration, stating that while plug-and-play simplicity is tempting, Nuklai prioritizes auditability and consistency, sometimes opting for custom connectors over generic pipelines to avoid data distortion.

"We've chosen to prioritize consistency and auditability," he said. "Nexus doesn’t alter the underlying data or make assumptions about its truth. What we do enforce is traceability."

Also Read: Ethereum ETFs, Treasury Companies Now Hold Over $32B In ETH: Here’s What’s Driving The Frenzy

Nuklai's vision goes beyond software: it aims to build the connective infrastructure for AI-native economies.

The company's incentive model ties partner contributions to verifiable economic outcomes within the ecosystem.

Partners that try to silo their stack, de Vries warned, will miss out on the composability and shared trust that the ecosystem provides.

In practical terms, Nexus is already integrated with io.net's decentralized model hosting infrastructure, enabling users to run queries using hosted large language models without relying on centralized services like OpenAI.

"Developers gain instant access to powerful models without needing to manage infrastructure," said io.net's Tausif Ahmed.

While full-scale deployments are still ramping up, de Vries said Nexus is already live with early users and will expand further in coming months.

The priority, he said, remains clear: real-world orchestration of tools, context-aware AI agents, and data systems that don't hallucinate under pressure.

Read Next:

• 97% Of Bitcoin Wallets Are Now In Profit -- What Does It Mean For Your Portfolio?

Image: Shutterstock