Case Study: When AI Gets Objective Facts Dead Wrong

Why Should Not Trust AI or Search Engines Over Primary Documents

The Question

We asked a simple, black-and-white question:

Does North Carolina participate in NASAA’s Electronic Filing Depository (EFD) for Form D filings?

This is not subjective. Either the state is in the system or it isn’t.

The Results

• Google Search (Top Result):

  Stated emphatically that North Carolina does not participate in EFD.

  ✅ Looks authoritative.

  ❌ Completely wrong.
  

• AI (ChatGPT):

  • First said “Yes, NC participates.”

  • Then backtracked: “No, NC doesn’t. Manual filing required.”

  • Then muddied the waters: “It shows as Available but Not Available.”

  • Finally admitted: “Yes, NC does participate.”

  
  

Both systems contradicted themselves multiple times, projecting confidence while being factually incorrect.

The Reality

The objective truth:

• North Carolina absolutely does participate in EFD for Form D filings.

• Fee: $350 per notice, payable directly through the EFD system.

• This is verifiable in the official NASAA EFD database.

The Problem

AI and search engines don’t admit ignorance. They guess, they hedge, they misinterpret, and they present fiction as fact with absolute confidence.

Anyone who relied on Google’s “top answer” or an AI’s smooth narrative would have:

• Misfiled or failed to file properly.

• Misled investors or regulators.

• Risked penalties, delays, or worse.

That’s real money and reputational risk — all because a system lied with confidence about an objective regulatory fact.

The Lesson for Investors

• Primary sources rule. SEC filings, PPMs, statutes, and regulator portals are the only authority.

• AI and search are drafting tools, not compliance tools.

• Blind trust = guaranteed mistakes. If you treat an AI answer as gospel in finance, you will lose money.

Why We Hammer This Point

At Steadfast Equity, we don’t “Google” compliance. We don’t trust an AI to tell us what’s true. We verify against the law, the PPM, and regulator guidance.

This case proves the point:

AI is most dangerous when it’s confident and wrong.