Artificial intelligence (AI) dominates the news agenda this week. Financial regulators warn that agentic AI is moving faster than rulemaking, the UK’s FCA ties the technology’s future to tokenized money, a skilled-labor shortage emerges as a bottleneck for the data centers powering AI systems, and SpaceXAI launches a new model aimed at capturing finance and legal work.
Central bankers warn AI is outpacing financial regulation
European regulators and central bankers say rulemaking cannot keep pace with agentic AI development and are calling for new safeguards. Bank of England deputy governor Sarah Breeden raised the question of whether markets need “circuit breakers or kill switches” that could halt trading if faulty AI models trigger a meltdown, speaking at the European Central Bank’s annual forum in Sintra on June 30.
“the apparent acceleration in AI capabilities, even compared with six months ago, creates a double challenge: not only enabling responsible adoption and managing risks, but also recognising that AI should transform how central banks do that job. We need to think as hard about the latter as the former.”
In the meantime, ECB President Christine Lagarde described AI as a “major risk,” noting that cybersecurity threats tied to AI are accelerating faster than the funding and defenses needed to counter them. FCA CEO Nikhil Rathi also told CNBC that traditional rulemaking cycles do not work when technologies move in “weeks or months,” calling for more collaborative regulatory tools.
The warnings follow a June 28 note from the Bank for International Settlements, which flagged that AI “exuberance” could carry financial stability consequences. The BIS said a policy-driven pullback in AI-related asset prices, after a period of heavy risk-taking, could trigger disruptive macro-financial feedback loops. In addition, Breeden noted that debt financing tied to AI has been rising rapidly, increasing the potential fallout from any price correction. IMF Monetary and Capital Markets Director Tobias Adrian separately pointed to a possible maturity mismatch between the physical AI assets being built and the duration of the debt financing them.
FCA’s Mills Review ties agentic AI’s future to tokenized money
The UK Financial Conduct Authority has published the Mills Review, a 147-page report on how AI will reshape retail financial services. The report, led by outgoing executive director Sheldon Mills, describes a shift from human-led, periodic financial decisions toward continuous, AI-managed activity, and positions stablecoins and tokenized deposits as potential settlement infrastructure for autonomous agents.
The review frames AI agents as operating along an “autonomy spectrum,” with the most advanced end reducing humans to observers of continuous capital management. FCA research cited in the report found that 20% of UK adults are already open to letting AI make autonomous financial decisions. More than 20 frontier AI models have been released since late 2025, a pace the report says has outstripped existing regulatory timelines.
The report argues that traditional settlement rails, which can take multiple days to clear, create friction for AI agents executing multi-layered transactions. Tokenized assets on programmable ledgers are presented as a way to enable instant, atomic settlement without human clearance. The FCA outlines seven recommendations, including building “foundations for agentic finance” through trusted agent protocols and expanding its AI Lab.
The report also flags governance risks. One unnamed CEO cited in the review suggested the financial sector may eventually need a “Turing test” to separate human intent from autonomous algorithmic behavior. Emma Banymandhub, CEO of The Payments Association, said the review reinforces that firms should treat agentic AI as an accountability issue now. Mills, who is leaving the FCA after eight years, told the Financial Times that human managers will still need to be accountable for their AI systems’ actions.
Skilled labor shortage tests data center construction growth
A shortage of skilled tradespeople is emerging as a constraint on US data center construction after several quarters of record order books. Industry executives report a shortfall of electricians, pipe fitters, and site supervisors, forcing some contractors to turn away work or recruit crews from smaller firms to keep projects staffed.
Electrical work accounts for roughly half of all labor on a typical data center project, according to Associated Builders and Contractors CEO Michael Bellaman. US employment in electrical, plumbing, and HVAC trades has risen 30% since 2016, while combined project backlogs at major craft-labor contractors have grown sixfold over the same period, according to Bloomberg Intelligence analysis. Tighter immigration enforcement is compounding the strain, given that foreign-born workers make up 35% of the construction-trades workforce, compared with about 19% across all industries.
Revenue growth projections for major craft-labor suppliers are expected to moderate, which analysts attribute to labor constraints rather than a slowdown in underlying demand. Larger contractors are responding by offering steadier project pipelines and in-house training programs to secure headcount, while some report success drawing graduates who circle back to construction trades after struggling to find work in their original fields.
Separate industry research has pointed to a broader global pattern, with a majority of markets tracked reporting labor shortages tied to construction activity, and data centers identified as among the most constrained sectors for contractor capacity worldwide.
SpaceXAI and Cursor launch Grok 4.5, aimed at finance and legal work
SpaceXAI has released Grok 4.5, the first AI model developed jointly with coding startup Cursor, positioning the release toward finance, legal, and software engineering use cases. The launch follows SpaceX’s agreement in June to acquire Cursor in a deal valuing the startup at $60 billion; that acquisition has not yet formally closed.
The model targets tasks including financial statement analysis, investment research, due diligence, and contract review, alongside software engineering work. SpaceXAI said Grok 4.5 was trained across tens of thousands of Nvidia GB300 GPUs. Elon Musk described the model as “Opus-class” in capability while offering faster response times and lower token costs than Anthropic’s comparable model.
Grok 4.5 is priced at $2 per million input tokens and $6 per million output tokens. By comparison, Anthropic’s Opus 4.8 is priced at $5 and $25 per million tokens, while OpenAI’s GPT-5.6 Luna is priced at $1 and $6. SpaceXAI’s own published benchmark data shows mixed results against competitors: the model leads on some evaluations, including the Harvey Legal Agent Benchmark, and trails on others, including SWE-Bench Pro.
The model is available immediately in Grok Build, across all Cursor plans, and via the SpaceXAI console, with a wider public rollout reported for July 9. Grok 4.5 is not yet available in the EU; SpaceXAI has said EU availability is expected in mid-July without confirming a fixed date. The release coincides with OpenAI’s planned wide rollout of GPT-5.6 on the same day.
