Here’s what’s on our plate today:

• 🧪 Free, fluent & unaccountable: inside the new money classroom.

• 📰 Altman ready to decelerate; AI stocks slide on capex; AI surveillance lampposts.

• 🛠️ Three tools worth trying: MoneyHelper, Khanmigo, FCA Mills Review.

• 🗳️ Poll: how should we handle AI as a money teacher?

Let’s dive in. No floaties needed…

The Laboratory

TL;DR

• AI is the new tutor: FCA research finds 16% of UK consumers already use AI for personal finance, mostly to explain concepts and suggest actions, placing it where advisers and parents once sat.

• Trust is split: 75% of users trust AI to explain clearly, but only 23% trust it to avoid misleading them, and familiarity erodes the instinct to double-check.

• No safety net: Only 40% of consumers know there is no formal recourse if chatbot advice goes wrong, and regular AI users are the most mistaken.

• Regulator's concession: The Mills Review proposes a public-interest AI capability service, accepting that teaching has moved from people to systems.

• Stakes: A generation with record-low financial literacy is learning money from a fluent, unaccountable machine, and nobody answers for the lesson.

Free, fluent & unaccountable: inside the new money classroom

Since the dawn of human civilization, the ability to transfer knowledge about life, morals, technology, and social values has allowed societies to thrive long after the original proponents of an idea have died. For this transfer to occur, there has always been some form of education system. Students are exposed to an idea, try to implement it in the real world, learn from their mistakes, and then add to the original idea itself.

The system has worked for generations because each generation is taught to value the work of its predecessors without blindly relying on it. This, in turn, cultivates critical thinking, allowing students to ask questions with a particular assurance: the answers come from people who have spent decades studying a subject, and from institutions that stand behind those people when the answers turn out to be wrong.

In 2026, that equation is facing a new challenge, one that is the product of the technology's own extraordinary growth. Artificial intelligence, and the chatbots it powers, is changing the way information passes from one generation to the next. For the first time in history, the blackboard at the front of the classroom can rewrite the lesson on its own, and it is not always accurate.

The regulator walks into the classroom

Nowhere is that shift easier to measure than in how people learn about money. On July 6, 2026, the UK's Financial Conduct Authority published the Mills Review, a study of how AI will reshape retail financial services by 2030, and the first exercise of its kind attempted by a financial regulator anywhere. Underneath the review sits a survey of more than 5,000 UK adults who hold an everyday bank account, run in April 2026, which measured how people actually use AI with their money rather than how they expect to someday.

16% already use an AI tool for at least one personal finance task. Use runs highest for investing, debt management, and tax planning, the corners of financial life where people have always felt they needed an expert.

The troubling aspect of the findings goes beyond how many people use it and emerges when one looks at what they use it for. Among those already asking AI about money, 72% ask it to summarize, explain, or simplify information, and 61% ask it to suggest what they should do. Almost nobody is handing it the keys to an account. They are asking it to teach, which means the tool has taken up residence at the exact point in the chain where a parent, an adviser, or a more experienced colleague used to sit.

A teacher trusted for delivery, not correctness

The survey then measured the shape of that trust, and the shape is unlike anything a classroom has produced before. Among people who have used AI, 75% trust it to explain things clearly, and 71% trust it to give useful guidance. Only 27% trust it to admit when it does not know something, and just 23% trust it to avoid giving misleading information.

Read those numbers together, and a strange portrait emerges. People believe the explanation is excellent and simultaneously suspect it may not be true. No previous generation of students has been asked to hold both beliefs about a teacher at once, because the old system never separated the quality of the teaching from the reliability of what was taught. The institution behind the teacher fused the two.

The friction was the feature

There is a genuinely strong case for the new teacher, and it deserves its full weight. Financial knowledge has always traveled through an unusually weak chain, because human advice was rationed by price and basic questions carried a tax of embarrassment. A tool that is free, endlessly patient, and available at two in the morning is doing work no institution was doing. The approach can clearly work: a randomized Harvard trial found that students using a purpose-built AI tutor learned more than twice as much as peers in a well-run classroom, in less time.

But that trial contains the tension of this whole story. The Harvard tutor held the correct answers in advance, released help one step at a time, and had students attempt the work before explaining anything. It was engineered to create friction because that’s where learning happens.

But unlike a controlled response system, consumer chatbots are engineered to do the opposite: they remove effort, supply the answer immediately, and are tuned to keep the user satisfied.

Most people still do the checking themselves. Among those who used AI while shopping for shares or funds, 78% compared its advice against other sources before acting. The trouble is that the checking erodes with familiarity. The FCA's researchers separated regular users who trust AI both to avoid being misled and to admit ignorance, and found that 63% of that group treat it as a reliable source of financial advice, compared with 27% of their more skeptical peers.

A separate survey of 319 knowledge workers found that higher confidence in an AI tool predicted less critical thinking about its output. And researchers at MIT and Penn State found that over long conversations, personalization makes models steadily more agreeable, less willing to tell a user they are wrong. A teacher who agrees with you more the longer you study under them is running the education system in reverse.

A classroom with no complaints office

The deeper problem sits outside the conversation entirely. If a regulated adviser gives you bad advice, a machinery of consequences switches on: professional obligations, a complaints process, an ombudsman. If a general-purpose chatbot does the same, none of that exists. Only 40% of consumers correctly understand this. Worse, the people most exposed are the most mistaken: 41% of those who use AI for their finances wrongly believe some route to redress exists, roughly double the rate among consumers overall.

The timing compounds this problem, since the youth with the least experience are the ones leaning hardest on chatbots for information. Institutions are only beginning to respond to this shift, and regulators are discovering that consumers have moved faster than the rulebooks governing financial advice.

On April 6, 2026, the FCA's new 'targeted support' rules took effect, letting firms offer suggestions to groups of similar customers, a regime built over four years to fill the gap between plain information and expensive personal advice. The survey shows that consumers had already filled that gap themselves using a tool the regime does not cover. And the students arriving in this classroom know less than any previous cohort. The 2026 Personal Finance Index recorded the lowest US financial literacy score in its 10-year history, with Gen Z answering 38% of questions correctly, and found AI use for personal finance highest among that same generation. The people with the least knowledge are leaning hardest on the teacher, who is least answerable for it.

The state considers hiring its own teacher

The Mills Review's most revealing recommendation is its seventh: the FCA should consider building a trusted, public-interest, AI-enabled financial capability service. Having found that millions of people already learn about money from an unaccountable machine, the regulator's answer is to contemplate an accountable one. That is a sensible response and a concession. It accepts that the teaching function has permanently moved from people to systems, and narrows the remaining question to who should own the system.

The lessons learned in the financial sector, especially in how people educate themselves, have far-reaching implications. Every civilization has eventually built institutions around its teachers, precisely because the transfer of knowledge was too important to leave to whoever spoke most fluently. The choice now forming is whether societies rebuild those institutions around the new blackboard or decide that fluency is enough.

The old chain endured because every generation could interrogate the one before it, and because somewhere behind every answer stood someone who had to answer for it. The new teacher is more patient than any human who has ever held the role, cheaper than any institution, and available to anyone who asks. What remains unknown is what becomes of a generation whose most trusted teacher never has to be right.

Thursday Poll

3 Things Worth Trying

• MoneyHelper: The UK's free, government-backed money guidance service, a useful accountable benchmark against whatever a chatbot tells you.

• Khanmigo: Khan Academy's AI tutor built to create friction and make you attempt the work first, the opposite design to a satisfaction-tuned chatbot.

• FCA Mills Review: The full regulator report on how AI will reshape retail finance by 2030, worth skimming to see where the guardrails are heading.

Quick Bits, No Fluff

• Altman ready to decelerate: Sam Altman signaled he's open to slowing the AI race, a notable shift in tone from a CEO who has spent years pushing for speed.

• AI stocks slide on capex: AI stocks fell after Google's soaring capital spending spooked investors, reviving doubts about whether the infrastructure boom will pay off.

• AI surveillance lampposts: Privacy advocates are alarmed by AI-enabled "smart" lampposts, warning that the street furniture could quietly become a mass surveillance network.

The Toolkit

• Leonardo AI: AI image and video generator with fine-grained creative controls, built for designers, marketers, and game studios who need consistent style at scale.

• Modal: Serverless cloud for running Python and AI workloads, lets you spin up GPUs in seconds without touching infrastructure.

• Quillbot: AI writing assistant that paraphrases, summarizes, and rewrites text on demand, useful for tightening drafts or escaping your own voice.

Rate This Edition