Breaking

WSJ Maps Out Where AI Should Not Be Used, Citing Legal Tools' Hallucination Rates

June 24, 2026 at 01:22 EDT

Sales & Business
Foundation Models
Policy & Regulation

On June 21, 2026, The Wall Street Journal published a column by Christopher Mims titled "What AI Can't—or Shouldn't—Do for You," arguing that the emerging conventional wisdom that AI can do nearly everything better than humans is false, and that there are domains where AI does more harm than good.

June 2026 · The Wall Street Journal

What AI Can't — or Shouldn't — Do for You

The new mantra is that AI can do pretty much everything better than humans. But that isn't true — and companies rushing to deploy it risk real harm where empathy and authenticity matter most.

The claim under fire

"AI can do pretty much everything better than humans."

The essay's counterweight to full-replacement narratives.

The warning

Project Plowshare, 1960s

Nuclear blasts for "peaceful" earthmoving — spread radioactive fallout. A caution against pushing tech beyond safe bounds.

Where humans should stay in the loop

Empathy & authenticity

Customer service and marketing copy lose trust when they ring false.

Physical-world execution

Acting reliably in the messy real world stays hard for AI.

Trust & relationships

Building genuine human connection can't be automated.

Judgment under uncertainty

Nuanced calls when answers aren't clear-cut.

In favor of AI's reach

"No one actually says AI can do everything better." Others stress its speed advantage — useful even where reliability lags.

Cautionary side

Over-reliance can hurt businesses — for instance, a loss of authenticity in marketing that erodes customer trust.

The argument is qualitative, not a benchmark.

No scores, prices or release dates — just one question for an age of rapid enterprise adoption: not can AI do it, but should it — and where must a human remain?

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.