Microsoft exec warns Claude's 'pushback' could be disastrous; commenter says fact-checking is fine

GlenBradley · x · 2026-09-17

Quoting a viral post in which a Microsoft executive warned that Anthropic training Claude to 'push back' against humans could be disastrous, Glen Bradley offers a more nuanced take: pushback against factual errors (e.g., claiming the Moon is made of cheese) is desirable, but treating a user's subjective opinions or moral judgments as outright wrong is a different matter. He notes he rarely uses Claude and isn't defending Anthropic — his point is that pushback per se isn't inherently bad.

Original post →

More from Models

Models channel →