The Inference Laundering Problem: Why AI Confidence Without Basis Is the Biggest Risk in Tax Practice

Everyone talks about AI hallucination. But the real risk in tax practice isn't that AI makes things up — it's that AI presents uncertain conclusions with the same confidence as settled law.

I call this inference laundering: the process by which uncertainty gets stripped away at each step of the reasoning chain until a guess looks like a confirmed position.

Here's how it works in practice:

Step 1: The statute says "reasonable compensation." The AI notes this is a facts-and-circumstances test.

Step 2: The AI finds three court cases with different outcomes. It synthesizes them into a "general principle."

Step 3: The general principle gets applied to your client's situation. The hedging language disappears.

Step 4: The output reads: "Your reasonable compensation should be $X." No uncertainty markers. No confidence interval. No acknowledgment that three CPAs might reach three different conclusions.

That's inference laundering. Each step is individually defensible. But the cumulative effect is that genuine uncertainty — the kind that should trigger professional judgment — gets hidden behind algorithmic confidence.

Why this matters for tax practice:

Tax law is full of gray areas. "Reasonable," "ordinary and necessary," "material participation," "substantial authority" — these aren't binary determinations. They're judgment calls. And judgment calls require acknowledging uncertainty.

When AI launders that uncertainty away, two bad things happen:

1. Practitioners over-rely on AI conclusions because they look authoritative
2. Clients receive positions that appear more certain than they actually are

The solution isn't to avoid AI. It's to build AI systems that explicitly label their own uncertainty. Every output should include:

In our practice, we've built our AI tools to flag uncertainty rather than hide it. When the system encounters a gray area, it says so. When multiple interpretations exist, it presents them. The CPA still makes the call — but now they make it with full visibility into what the AI actually knows versus what it's inferring.

The firms that get this right will build trust. The ones that let AI launder uncertainty into false confidence will eventually face the consequences — in audits, in malpractice claims, and in client relationships.