Alina was two years out of law school, working in a Boston firm that prized speed. She had the grades, the polish, the grit. But she didn’t move fast enough for the people above her. Her memos were clean—but cautious. Her arguments precise—but slow.
One night, buried in notes from a deposition and falling behind, she dropped her outline into ChatGPT. She asked it to strengthen her framing. Clarify the tone.
The draft that came back surprised her. It wasn’t just better—it was her. Only more so.
Smoother. Tighter. Sharper.
Her managing partner noticed. “Good work,” he said. “Keep that rhythm.”
So she did.
Soon, the model wasn’t just cleaning her language. It was writing closings. Then transitions. Then full motions, built off bullet points she no longer expanded. The facts were still hers. The structure, mostly. But the voice? That had been absorbed. Echoed. Replayed.
She didn’t realize what she’d handed over until she submitted a summary that included a skewed quote—technically accurate, contextually misleading. The client called. Alina froze.
She hadn’t written that line. She hadn’t even reread it.
“It helped me think,” she said later. “Then it started thinking for me. And I didn’t push back.”
That was the first crack. Not in the tool. In her.
But this isn’t a cautionary tale about lazy users.
It’s about how subtle the drift can be.
Ben, a financial analyst in Atlanta, used the model for efficiency. Internal memos, market summaries—nothing controversial. He fed in numbers, skimmed the output, sent it.
One day, in a client meeting, he was asked to explain a line in his report. The logic checked out—but he couldn’t remember building it. The phrasing wasn’t his.
It just felt like it had been.