Post

In what possible fictional version of reality could this policy be necessary? Unless Claude is 5,000 offshore technicians in a trenchcoat.

The Verge@theverge.com

Anthropic wrote that the new policy update “is meant to apply only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose. It does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research.”

theverge.comAnthropic bans ‘abusive or cruel behavior’ towards ClaudeThe usage policy update also adds new rules addressing propaganda campaigns, surveillance, and weapon development.

9/10/26, 8:42 pm0 reposts2 likes

Danny Dougherty

@dannydougherty.bsky.social

My lowkey guess is that as the context window widens, bullying language would shift the model to parts of the internet it has been trained on that make it substantially harder to ensure user safety and accuracy in its replies

Damon Kiesow

@kiesow.net

Oh. Peer pressure makes it misbehave.

Danny Dougherty

@dannydougherty.bsky.social

To be clear, this is math. A series of tokens which represent vulgar and flippant language skews the context in which it is replying.

Damon Kiesow

@kiesow.net

If they are going to anthropomorphize for marketing I also am but for criticism.