Snowflake CTO: UK gov must close the data-literacy gap to lead in AI

TL;DR:

• Snowflake field CTO Fawad Qureshi argues UK government will not lead in AI until it closes a clear data-literacy gap between Whitehall and the commercial sector.
• Public trust is fragile and conditional: DSIT’s own research into the proposed National Data Library shows citizens default to concerns about “undisclosed intentions” without clear governance.
• The talent and capability constraint — heavy reliance on contractors, constrained pay, structural inefficiencies — is the binding issue: “You cannot regulate what you do not understand.”

Behind every UK AI strategy paper sits an awkward truth: government’s understanding of, use of, and governance over its own data lags the commercial sector. That is the framing Snowflake field CTO Fawad Qureshi sets out in a wide-ranging commentary published this week, arguing that AI leadership ambitions cannot be met until that basic gap is closed. “There is a knowledge gap between the public and the commercial sector,” Qureshi said. “What we assume is common sense or straightforward is not always so common in government.”

Bias starts upstream — and trust is fragile

Qureshi’s core technical argument is that data is never neutral. “We are all humans. We all have our own biases. Data is capturing those behaviours — it’s not neutral. The risk is that we amplify those biases and put people at systematic disadvantage.” The implication for UK departments is that AI investment must shift upstream: provenance, diversity, context and collection environment matter more than the model layer. His refugee-data example is sharp: “Why should someone fleeing across borders trust you enough to be honest in a survey? For them, you are an adversary.”

The trust dimension is reinforced by the government’s own evidence base. The Public Attitudes to Data and AI Tracker Survey has consistently shown NHS England as one of the UK’s most trusted data-handling institutions — but trust drops sharply where oversight is weaker or commercial involvement is visible. Recent DSIT research into the proposed National Data Library found that citizens default to concerns about “undisclosed intentions” unless governance, transparency and explanation are clear and concrete. “When you collect data for one purpose, use it for that purpose,” Qureshi said. “Don’t quietly use it for something else. You will break trust — and once it’s gone, it’s gone.”

The capability question Westminster keeps deferring

The harder argument is about talent. The current government model — constrained pay scales, heavy contractor reliance, structural inefficiencies — makes it difficult to attract the people who can hold the AI policy conversation at the technical level it requires. “When the best people are not in government, you end up outsourcing policy work you cannot fully understand,” Qureshi said. The point lands with particular force in a week when HMRC has signed a £175 million Quantexa partnership and the Bank of England, FCA and HM Treasury have issued a frontier-AI cyber statement that depends on regulated firms — and supervisors — being technically literate enough to act on it.

Looking forward

Qureshi’s prescription for UK government is to treat data as a strategic asset, protect public trust as a core capability, design for long-term societal impact, embed accountability into automated systems, build verification into the fabric of digital services, and create an environment where top talent wants to work in government. None of these are quick wins. But the implicit warning — that AI strategy without data and capability foundations becomes either performative or quietly damaging — is one this week’s UK AI investment headlines do not engage with directly. “Trust is earned in drops and lost in buckets,” Qureshi said. “One misuse of data, and people remember it for decades.”