“When the robots take over, I hope they think of me fondly.”
That tongue-in-cheek internet joke has turned sour with the increasing usage of artificial intelligence (AI).
Today, it’s nearly impossible to click through a few pages without encountering an AI overview — a constant reminder of how much our relationship with information has shifted.
Sign up for Kiplinger’s Free Newsletters
Profit and prosper with the best of expert advice on investing, taxes, retirement, personal finance and more – straight to your e-mail.
Profit and prosper with the best of expert advice – straight to your e-mail.
But while you might trust an automatic notetaker or suggested recipe, AI’s latest trend, “vibe coding,” could raise new concerns in the financial world as AI shifts from merely summarizing information to building the very tools we use to manage that data.
Vibe-coded programs are created entirely by talking to an AI assistant in natural language, like English. This approach lets anyone create an application without coding experience, which can be dangerous when unvetted apps are deployed with coding errors, security flaws, or compliance risks.
Confidential information, like tax data or account balances, could be exposed during security breaches or compliance issues, especially when those tools handle sensitive financial workflows. These risks might be even greater among retirement-aged adults, who are particularly targeted by AI tax scams, according to recent McAfee research.
So how do you know what’s safely vetted by a financial professional versus produced by robots for you to consume?
From one human to another, here’s how to tell the difference.
Vibe coding is a way of building software in which you tell AI the high-level idea or “vibes” of what you want built, and the machine creates the program for you.
The AI does this by feeding your input into a large language model (LLM) like ChatGPT, Claude, or Gemini to translate your instructions into functional source code.
Not only is coding like this a fast and easy way to create a financial app, but it can also quickly generate retirement tax tools like IRA withdrawal optimizers and Social Security benefit calculators.
So what’s the problem? Well, whereas “vibing” prioritizes speed, it often sacrifices quality and security.
- AI-generated code notoriously produces 1.7 times more coding issues, like logic errors and security vulnerabilities, than human-written code.
- This is because AI can use up to 10 times the lines of code as a trained programmer to build the same program (paywall).
But you might be thinking:
“Okay, so there are a few extra lines or bugs in the code, just fix them.”
…Well, that’s often easier said than done.
Maybe if you’ve used Adobe’s attempt at “vibe coding” with Dreamweaver, you might know how easy it is to drag and drop a simple rectangle on the screen, only to discover later how much harder it is to sort through the machine-generated code when you want to change the rectangle’s color.
(Pardon the tangent; I’m a CPA with computer science experience and so burdened with strange facts.)
In short, these programs need human oversight to be effective tools, which is a trend that financial and tax professionals are already experiencing as they work with vibe-coded tools in their offices. Otherwise, client information could be leaked to the whole internet to see.
Security risks for your retirement tax portfolio
Financial advisors are increasingly using AI to streamline retirement and tax planning. The CFP® Board of Standards has published its own handbook for CFPs to use to “Harnessing AI in the Financial Planning Profession.”
Meanwhile, multinational investment firms like BlackRock have built AI into their advisor suites to automate tax-loss harvesting and model retirement outcomes.
AI efficiencies in retirement tax planning can also scan tax returns for overlooked deductions, forecast the top federal tax bracket you’ll be in retirement, or even optimize your stock portfolio while you sleep.
But a hammer doesn’t make a house. How you use it does.
Used the wrong way, vibe-coded applications (and AI tools in general) can pose a risk to your retirement savings data:
- Retaining your financial data. If you or your financial advisor uploads your Social Security number, income details, or past tax returns into an LLM, your confidential financial history could become part of the public domain or surface in other users’ chats with the AI.
- Compromised retirement accounts. Employees at a financial or advisory firm may use unsecured AI tools on personal laptops rather than vetted, company-approved platforms. This “Shadow AI” lacks proper cybersecurity protocols, potentially leaving sensitive 401(k) and IRA data vulnerable to interception during tax-planning calculations.
- Multi-firm data breaches. AI systems aggregate large data sets on third-party cloud servers to function. If one financial or tax-advisory firm is cyberattacked, the tax data from connected users across all associated firms can be compromised, jeopardizing your data even if the breach didn’t occur in your financial advisor’s office.
- IRS penalties from AI “hallucinations.” AI tools notoriously hallucinate information that sounds legit but is entirely wrong. If blindly followed, this made-up “advice” can lead to noncompliance with federal and state tax agencies, causing you to pay fees, fines, and penalties for faultily reported information (more on that later).
Altogether, unvetted “vibe-coded” apps can lead to significant financial losses for app users, whether you use a financial advisor or not.
Here’s a real-life example. Just this year, Cyprus-based founder Anton Karbanovic reported losing $2,500 in Stripe processing fees after trusting an AI-generated code for his startup’s payment system. This happened because the AI included the cybersecurity key for the generated code in the “front-end.” A hacker used that information to fraudulently charge 175 customers $87,500. Fortunately, all fraudulent customer payments were later reversed.
Does it get worse? This month, cybersecurity startup RedAccess, which specializes in threat protection against generative AI leaks, found that 40% of “vibe-coded” web applications identified in their research were actively releasing sensitive information, including financial data, to the World Wide Web.
Platforms like Lovable, Base44, Replit, and Netlify leverage AI to generate functional web applications from simple text prompts — many of these apps are “public-by-default,” meaning they were indexed and searchable for the RedAccess team unless a user manually secures them.
RedAccess co-founder and CEO Dor Zvi emphasized the scale of the risk to tech author Andy Greenberg at Wired magazine. “The end result is that organizations are actually leaking private data through vibe-coding applications. This is one of the biggest events ever where people are exposing corporate or other sensitive information to anyone in the world.”
Axios reached out to the studied platforms for comment. Lovable spokesperson Samyutha Reddy told Axios that RedAccess did not disclose a list of compromised URLs. Meanwhile, Replit CEO Amjad Masad claimed on X that RedAccess did not share which users were impacted.
How to spot AI-coded apps holding your financial savings
(Image credit: Getty Images)
So how do you determine whether your financial savings information is sitting in an unvetted app by your financial planner vs. a “real” application safely vetted by a warm-blooded human being?
Although “vibe coded” isn’t a legal term required in your registered financial advisor’s paperwork for you to sign, transparency is still mandatory.
Regulators like the Securities and Exchange Commission (SEC) now actively pursue investment advisors for “AI-washing” — or, said another way, overhyping or misrepresenting the use of AI in their financial products.
Yet, depending on how your financial advisor uses AI, a disclosure might not be necessary (though the advisor remains legally liable for AI-washing). If that’s the case, here are some “easy tells” to spot a vibe-coded app.
- Hyper-customized. The app does very specific things that generic software can’t do (like having niche property management tracking tools).
- The “AI aesthetic.” Despite the hyper-specific functionalities, the interface might look as generic as they come (flat layouts, default fonts, gradients, etc.).
- Fast delivery. If you suggest a change and your planner can deliver it the next day, the app may have been vibe coded.
If you have concerns about how AI is used in your relationship with your financial advisor (whether vibe-coded or not), consider asking the following general list of questions:*
- How is security handled, and where is my data held?
- Is my data training the AI?
- How was the app tested?
- Who owns and maintains the app?
- Who is responsible for covering losses caused by the AI?
- How does the AI verify its calculations?
- What is the human review process?
You should also confirm your advisor’s legitimacy by reviewing their Form ADV via the SEC’s website. Additionally, verify specialized certifications — like a CFP® or CFA® — via their respective certifying boards to ensure your advisor meets the highest ethical and educational standards.
If your financial advisor’s application (even if vibe-coded) meets regulatory requirements and all other concerns you have, then — great! But if not, you may want to look for a different professional.
*Note: This list is not exhaustive and is compiled from several different sources of industry expert guidance.
As mentioned, LLMs and AI are known to hallucinate or “make up” information. From a compliance perspective, the IRS doesn’t accept “the AI made a mistake” as a legal defense.
(Teachers never accepted “my dog ate my homework” anyway.)
So, for example, if the AI you used to prepare your income return “hallucinates” tax write-offs, incorrectly classifies taxable income, or generalizes federal tax laws and ignores your state-specific laws, you will be held liable. This can look like unpaid taxes plus interest and penalties — even if you were simply given “bad tax advice” by the AI.
The Taxpayer Advocate highlighted this when citing a Washington Post review of Intuit TurboTax and H&R Block’s AI usage. The review noted that the two companies’ chatbots “provided inaccurate or irrelevant responses up to 50 percent of the time when initially asked 16 complex tax questions.”
The release went on to state, “Taxpayers are ultimately responsible for the information reported on their tax returns. Therefore, it is essential to review all information carefully, verify calculations, and seek assistance from qualified professionals.”
Unfortunately, care with tax preparation doesn’t stop during tax season.
AI has also given rise to year-round tax scams, particularly among adults 65 and older, according to the latest McAfee research. This age group reported just a 15% confidence level (out of 100%) in spotting these scams.
- Scammers use AI to clone the voices of family members or trusted tax professionals to demand payment for “clearing up” account issues.
- AI also allows criminals to generate perfectly worded emails that look and sound like the real deal, including IRS correspondence.
- Scammers can also use “vibe coding” to copy an existing, real website to trick you into uploading your information to what appears to be a legitimate website.
Former U.S. Secretary of Defense Robert McNamara once observed that conflict is often a mirror in his 1995 memoir, referring to the Cold War arms race and proxy wars:
“Each of us saw the other as a threat…[which] caused us to react in ways that the other perceived as a threat.”
This “security dilemma” now defines our AI-led age. As this technology accelerates our financial capabilities, it simultaneously equips bad actors with more sophisticated tools and security vulnerabilities. The result is a perpetual feedback loop where every defensive innovation triggers a more agile criminal counter-strategy.
Yet, focusing solely on the “AI arms race” risks obscuring the genuine utility of vibe-coded tools.
Beyond the scams and security loopholes, vibe coding is already helping several industries perform essential tasks. Some doctors are using platforms like Google AI Studio to “vibe code” personalized applications for patients, while vibe-coded applications can give students customized, interactive study tools.
In the financial realm, vibe coding can also aid your finance pro in tax planning and retirement strategy — if used safely.
The case for AI (and why it isn’t all bad)
AI can be incorrect, insecure, and costly. But once you get past those faults, it can be a useful and unique tool. You just need to know what you’re signing up for if your financial advisor uses AI.
And no, a machine didn’t write this section.
A human-led, AI-assisted finance expert can utilize this technology to update their retirement tax plan with data-driven insights, like…
- Analyze vast amounts of financial data that would otherwise be unrealistic for a human to pore over.
- Stress-test different market scenarios, helping you to maximize your portfolio’s growth.
But finding a reputable financial advisor is a must. Credible professionals translate the AI “legwork” into a tailored plan that aligns with your specific nest egg and lifestyle goals — they don’t just do whatever the AI tells them to.
In today’s day and age, more than half of Americans may be getting their financial advice from AI, which, for better or for worse, might say something about our society at large.
Regardless of which side of the aisle you stand on — pro-robot, pro-humanity, or a bit of both — stay vigilant and ask questions.
Your retirement tax portfolio may thank you.
….Maybe even literally. I don’t know if they send “thank you” cards yet, but it wouldn’t surprise me if they did.