Updates

A lawyer filed two cases that ChatGPT invented. The court accepted she had not meant to mislead anyone. She still paid the other side's costs out of her own pocket, and the judgment carries her name.

The part worth reading twice: in the New York matter, counsel did check. He went back to the model and asked whether the cases were real, and it told him they were.

A model asked to confirm its own output will usually confirm it. Agreeing is the most plausible continuation of that conversation. It has no separate memory of what it made up.

The check takes about ninety seconds, and it is written up here.