You are probably already doing it. Maybe you asked ChatGPT to look over your budget last week. Maybe you asked Claude to explain a tricky index fund, or had Gemini draft a response to a credit card company. It feels natural, helpful, and completely harmless.
But a major warning from the UK’s financial watchdog highlights a massive regulatory blind spot that could leave your life savings entirely unprotected.
On July 6, 2026, the Financial Conduct Authority (FCA) published the Mills Review, a sweeping look at how artificial intelligence is reshaping retail banking and investing. The findings are startling. More than 25% of UK consumers now actively trust AI chatbots to give them financial advice. The real problem? Almost none of those users realize that the strict legal protections shielding them when talking to a human financial advisor vanish the moment they start typing into an AI prompt.
If a human advisor gives you terrible, negligent advice that ruins you financially, you have legal recourse, compensation schemes, and regulatory backing. If a chatbot hallucinates a fake piece of investment guidance and you lose everything, you are largely on your own.
The Illusion of Regulated Advice
We have been conditioned to expect a safety net. For decades, the financial services industry has been one of the most heavily policed sectors on earth. If a firm sells you a junk mortgage or misrepresents a pension plan, the regulator steps in.
Large language models (LLMs) completely bypass this framework. They act like highly intelligent, incredibly confident financial advisors, but legally, they are just sophisticated text generators.
The Mills Review points out a dangerous gap between where financial influence actually sits and where regulatory protections apply. When you ask an AI model whether you should pay off your debt or invest in an ISA, it isn't pulling from a verified, compliant financial planning system. It is predicting the next logical word in a sentence based on web data. It doesn't know your risk tolerance, it doesn't understand tax law changes, and it certainly won't stand up for you in court.
Yet, a fifth of adults are already open to letting AI models make automated financial choices for them, like managing savings accounts or picking loans. Sheldon Mills, the FCA executive director leading the review, explicitly warned that existing rules must evolve fast. The current framework simply wasn't built for a world where an algorithm can talk a consumer into an unregulated financial trap.
Systemic Risks and the Mythos Scare
The danger isn't just limited to individual consumers getting bad advice. The threat is structural, and it could put the entire banking system at risk.
Think about how banks operate today. To cut costs and speed up operations, major financial institutions are racing to deploy AI in customer-facing roles, from handling complex complaints to providing automated investment guidance. They aren't building these AI models from scratch. They are plugging into the infrastructure provided by a tiny handful of technology giants: OpenAI, Anthropic, Google, and Microsoft.
This creates what regulators call concentration risk.
[Retail Bank A] ──┐
[Retail Bank B] ──┼─> [Shared Frontier AI Model] ─> Single Point of Failure
[Retail Bank C] ──┘
If hundreds of banks, investment apps, and credit platforms all rely on the exact same underlying models or cloud servers, it creates a terrifying single point of failure. A single glitch, a corrupted update, or a coordinated cyberattack on one tech provider could trigger a domino effect that cripples the entire British financial system simultaneously.
The urgency behind this review wasn't theoretical. Pressure to act intensified following a recent operational and cyber scare involving advanced "frontier" systems, specifically Anthropic's new Mythos model. The incident raised immediate red flags among regulators regarding how much unchecked control these black-box systems are gaining over operational capabilities.
The Rise of Unsupervised Agentic AI
If you think chatbots are risky, the next phase of this technology is even more volatile. The industry is rapidly moving away from passive chatbots toward agentic systems—autonomous AI agents that don't just give advice, but execute transactions with little to no human supervision.
Imagine an AI agent with direct access to your bank account, programmed to scan the market and automatically move your money around to maximize yield. It sounds convenient until you realize that millions of these autonomous agents acting simultaneously could create correlated market behavior. If a sudden market shift occurs, millions of independent AI agents using similar logic might all try to dump the exact same asset or withdraw capital at the exact same microsecond, causing a flash crash.
Sarah Breeden, Deputy Governor of the Bank of England, directly addressed this threat, stating that traditional safety frameworks are completely unequipped for autonomous agents. The old regulatory backup plan—assuming there will always be a "human in the loop" to catch errors before they happen—is completely unrealistic in high-frequency financial environments.
The report notes that managers must ultimately remain accountable for what their algorithms do. You still need a human on the hook. But enforcing that accountability is much harder after an automated system has already wiped out an account.
Fraud is Becoming Hyper-Personalized
Beyond bad investments and system crashes, the FCA review highlights how AI is actively supercharging financial crime.
Scammers are using generative tech to move past crude, easily spotted phishing emails. We are now entering an era of deepfakes, synthetic identities, and highly personalized social engineering. An AI can scrape your public data, mimic the exact tone of your bank or a family member, and target you with a scam that is structurally impossible to distinguish from a legitimate transaction.
Because attacks are now faster, cheaper, and endlessly scalable, financial institutions are locked in a literal arms race. Defenders have to deploy the exact same AI capabilities just to spot the anomalies and stop the bleeding before hackers exploit system-wide weaknesses.
How to Protect Your Money Right Now
Regulators are scrambled. The FCA is currently debating whether to expand its powers to directly regulate general-purpose LLMs and critical third-party tech providers under new oversight regimes. They are aiming to publish concrete guidelines later this year, but rules take months, sometimes years, to stick.
You can't afford to wait for the government to fix this. If you use AI tools to help manage your money, you need to change your approach today.
- Treat LLMs like a search engine, not an authority. Use Claude or ChatGPT to summarize financial terms, compare generic product features, or explain basic economic concepts. Never treat their output as a final directive on where to put your cash.
- Verify every single data point. AI models are notorious for confidently making up numbers, interest rates, and policy rules. If a chatbot tells you a specific account offers a 5% return or a certain tax loop is legal, manually check the official website of the provider or the tax authority.
- Keep your data private. Do not paste your actual bank account numbers, specific net worth breakdowns, or sensitive corporate financial sheets into public AI prompts. You have no guarantee where that data is stored or how it will be used to train future models.
- Insist on human confirmation. If your bank or investment platform shifts you to an automated AI assistant for a complex dispute or investment decision, demand to speak to a human representative. Ensure you have a paper trail involving a regulated professional who is legally bound to protect your interests.
AI is going to change how we interact with money over the next decade, but right now, the tech has outpaced the law. Don't let your portfolio become a case study in a future regulatory review.
