AI News Digest: 23 March 2026

23 March 2026

A big day for UK AI Governance Stories today. Palantir has quietly extended its tentacles into the Financial Conduct Authority, the government's much-trumpeted OpenAI partnership has produced precisely nothing, and new data shows most British businesses still cannot measure a return on their AI spending. Meanwhile, the US chips-to-China smuggling scandal deepens, and Xiaomi's stealth trillion-parameter model continues to rattle the establishment.

Palantir Lands FCA Contract, Extending Its Reach Across the British State

The Miami-based data analytics giant has secured a deal with the Financial Conduct Authority, giving it access to terabytes of sensitive financial data from one of the world's most important centres of finance. It is the latest in a string of UK public sector contracts now worth more than half a billion pounds, following deals with the NHS, police, and military.

The FCA wants to use Palantir's AI-enabled systems to better detect financial crime and money laundering, aiming to reduce the number of investigative dead ends that currently consume regulatory resources. Campaign groups continue to criticise Palantir's work with US immigration enforcement and the Israel Defense Forces, but that has done nothing to slow its Whitehall expansion.

Our take: For UK businesses in financial services, this matters directly. Palantir's technology inside the FCA means AI-driven surveillance of market activity is now a reality, not a pilot. Firms should expect smarter, faster regulatory scrutiny and prepare their compliance frameworks accordingly. The broader pattern is clear: the British state is outsourcing its analytical capability to a single US firm at remarkable speed, and the debate about data Sovereignty is barely keeping up.

UK Government's OpenAI Partnership Has Produced Zero Trials in Eight Months

A Freedom of Information request has revealed that the Department for Science, Innovation and Technology has not conducted a single trial under the memorandum of understanding it signed with OpenAI. The partnership was announced with great fanfare as a vehicle for deploying advanced AI to "address society's greatest challenges" and reform public services.

DSIT pointed to a separate Ministry of Justice deal that gave civil servants access to ChatGPT, but critics argue that simply rolling out a chatbot hardly reflects the ambition of the original agreement. Tarek Nseir, founder of AI consultancy Valliance, called it "either a huge failure in execution or a failure of intent."

Our take: This is embarrassing but not surprising. Government AI partnerships tend to be long on announcements and short on delivery. The gap between political enthusiasm for AI and operational readiness remains enormous. For UK businesses working with public sector clients, the lesson is clear: do not assume government AI adoption will create near-term opportunities. The procurement and implementation machinery simply is not moving at the speed politicians promise.

78% of UK Firms Use AI, But Only 31% See Positive Returns

New research shows AI adoption is accelerating across UK businesses, with 78% now using the technology in some capacity and 85% among mid-sized organisations. However, just 31% report a positive return on investment. A further 18% say their AI initiatives have outright failed to deliver, while 16% say it is still too early to tell.

The core problem is strategic clarity. Only 41% of organisations using AI have a clear understanding of what success looks like, suggesting rapid uptake has outpaced organisational readiness.

Our take: This is the AI adoption gap we talk about constantly. Businesses are buying tools before defining problems. The 47% who cannot articulate measurable goals for their AI investments are effectively spending money on technology hope. If your organisation is in that bracket, stop deploying and start planning. Define what success looks like before you spend another penny on licences.

Supermicro Co-Founder Charged with Smuggling $2.5 Billion in AI Chips to China

The US Department of Justice has charged Yih-Shyan "Wally" Liaw, co-founder of Super Micro Computer, along with two others, with conspiring to smuggle billions of dollars' worth of Nvidia-powered servers to China in breach of export restrictions. Liaw has resigned from the Supermicro board. The company's shares plunged over 30% following the charges, though Supermicro itself has not been directly implicated.

Our take: This is the most significant AI export control enforcement action yet and sends a clear signal about the seriousness of the US-China technology cold war. For UK businesses in the AI supply chain, the compliance stakes just got considerably higher. If you are reselling, integrating, or distributing advanced AI hardware, review your export control processes now. The era of informal enforcement is over.

Xiaomi's Stealth Trillion-Parameter Model Revealed as MiMo-V2-Pro

The mystery model that appeared anonymously on OpenRouter under the codename "Hunter Alpha" in early March, quickly topping usage charts and processing over one trillion tokens, has been revealed as Xiaomi's MiMo-V2-Pro. The one-trillion parameter model delivers performance approaching GPT-5.2 and Claude Opus 4.6 at roughly one-sixth to one-seventh of the cost.

Our take: A smartphone manufacturer building a frontier AI model is a sign of just how fast this industry is commoditising. Xiaomi's approach of testing in production under a fake name before revealing itself is a new playbook that others will follow. For businesses evaluating AI providers, the message is: the cost of frontier-class capability is falling fast, and the assumption that only a handful of US labs can compete is increasingly outdated.

Europe's Power Grids Buckling Under AI Data Centre Demand

As data centre developers queue up to connect to power grids across Europe, network operators are experimenting with novel ways of clearing room for them. The AI race is creating an infrastructure bottleneck that could determine which countries lead in AI deployment and which fall behind.

Our take: The UK is directly affected here. Power availability is becoming the binding constraint on AI growth, not talent or capital. Businesses planning significant on-premise or co-located AI infrastructure should factor in grid capacity timelines that are measured in years, not months. This is also a major opportunity for UK energy companies and infrastructure investors who can move quickly.

Quick Hits

• Ecolab Makes $4.75 Billion Bet on AI Cooling. Ecolab has completed its $4.75 billion acquisition of CoolIT Systems, pivoting from water treatment to becoming a critical thermal management provider for AI data centres. As server rack densities push beyond what air cooling can handle, liquid cooling is becoming essential infrastructure. Another sign that the AI boom's biggest winners may be the picks-and-shovels suppliers, not the model builders. (Source: )
• LSE: When AI Sounds Right But Is Wrong. A timely piece from the London School of Economics arguing that AI fluency is not reliability. As LLMs become embedded in legal, financial, and clinical decision-making, the gap between sounding correct and being correct introduces structural risk. Essential reading for any business deploying AI in high-stakes contexts. (Source: )
• Jensen Huang Sits Down with Lex Fridman. The NVIDIA CEO discussed the company's path to a $4 trillion valuation and his vision for the AI revolution in a wide-ranging interview. Worth a listen for anyone tracking where AI infrastructure investment is heading next. (Source: )

Frequently Asked Questions

What does Palantir's FCA contract mean for UK financial services firms?

Palantir will use AI-powered analytics to help the FCA detect financial crime more effectively. This means UK financial services firms should expect smarter, more targeted regulatory scrutiny. Companies should review their compliance frameworks and ensure their own data and reporting systems can withstand AI-enhanced oversight.

Why are most UK businesses not seeing a return on AI investment?

Despite 78% adoption, only 31% of UK businesses report positive AI ROI. The primary reason is a lack of strategic clarity: fewer than half of organisations using AI have defined what success looks like. Businesses are deploying tools without clear objectives, measurable goals, or integration into existing workflows. Successful AI adoption requires defining the problem before selecting the technology.

What is Xiaomi's MiMo-V2-Pro and why does it matter?

MiMo-V2-Pro is a one-trillion parameter AI model built by Xiaomi that delivers performance close to leading US models like GPT-5.2 and Claude Opus 4.6 at a fraction of the cost. It matters because it shows that frontier AI capability is no longer exclusive to a small group of US labs. For businesses, this means the cost of accessing top-tier AI is falling rapidly, and competitive options are emerging from unexpected sources.