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 …

Full story available on Benzinga.com