> There were some regressions in the
model’s responses to user discussions about suicide and self-harm, and room for
improvement in some areas of child safety.
Someone had to make a decision somewhere this is an acceptable regression - wild. And then decide to write it down.