Millions of workers now rely on ChatGPT and other AI tools to edit emails, explain financial concepts, write reports, and even generate code. But as this quiet revolution reshapes how we work, it also raises serious questions: Are we sharing too much? Are we becoming dependent on a tool that might not only expose sensitive data but also fail us when we need it most?
A recent study from Indusface, an application security firm, is sounding the alarm.
According to their findings, over a third of U.S. adults who use AI tools consider themselves “dependent” on them for work-related tasks. Yet large language models (LLMs) like ChatGPT were found to fail 35% of finance-related questions, a troubling gap, especially considering how often people now ask these tools for money and business advice.
The Convenience Trap
It’s easy to see the appeal. AI is fast, always available, and surprisingly articulate. Many professionals within Fortune 500 companies are utilizing ChatGPT to enhance presentations, refine their communications, and generate new ideas. However, in our pursuit of productivity, we may be crossing a line, blurring the boundary between helpful assistance and excessive exposure.
Indusface reports that 11% of the data pasted into ChatGPT includes strictly confidential work information. That includes internal business strategies, financial reports, and in some cases, proprietary codebases. Unlike a coworker or consultant bound by confidentiality agreements, ChatGPT stores and learns from user inputs. That means your data could, intentionally or not, shape future responses, including those provided to someone else.
Don’t Feed the Machine: What Not To Share
Indusface recommends steering clear of inputting the following categories into any AI tool:
- Work files, including reports, internal strategy decks, and client presentations, often contain proprietary data. Even when anonymized, metadata or phrasing can still reveal more than intended.
- Passwords or access credentials: LLMs are not intended for use as password managers. Treating them as such opens the door to significant security breaches.
- Personal identifiers, such as names, home addresses, and photos, may seem innocuous but can be weaponized to commit fraud or create deepfakes.
- Company codebases: Developers increasingly use AI to debug or generate code, but doing so with proprietary source material could expose a business’s most valuable IP.
- Financial data: While ChatGPT can explain a Roth IRA or walk through budgeting basics, it’s not a CPA. Feeding it real figures and expecting a sound strategy is risky at best.
Is This a Tool or a Crutch?
At its best, AI can be a springboard—a way to brainstorm, double-check tone, or organize ideas. But at its worst, it becomes a crutch. That’s especially true in personal finance and business strategy, where precision matters and bad advice can cost real money.
So why are so many of us increasingly turning to a tool that’s explicitly not built to make decisions?
Part of it is habit. Part of it is the illusion of expertise. And part of it may be a growing discomfort with uncertainty, especially among younger professionals navigating complex career and financial systems without mentors or formal training.
Rethinking Our Relationship With AI
With International Passwordless Day approaching on June 23, now’s a good time to step back and reassess our digital habits. Are we prioritizing convenience over caution? Are we outsourcing too much of our decision-making, including career and financial matters, to a tool that’s designed to assist, not advise?
The lines between helpful and harmful aren’t always obvious. But here’s a simple rule of thumb: If you wouldn’t share it with a stranger, don’t share it with an AI.
And if you’re using ChatGPT as a substitute for real financial planning, legal advice, or strategic thinking, ask yourself: What happens when it gets something wrong? Will you even know?
Because while automation can speed up our workflows, it shouldn’t replace the critical thinking, professional judgment, and privacy protections that still matter—maybe more than ever.
